LEXICAL SEMANTICS AND SEMANTIC ANNOTATION James Pustejovsky with

  • Slides: 77
Download presentation
LEXICAL SEMANTICS AND SEMANTIC ANNOTATION James Pustejovsky (with additional slides from: Martha Palmer, Nianwen

LEXICAL SEMANTICS AND SEMANTIC ANNOTATION James Pustejovsky (with additional slides from: Martha Palmer, Nianwen Xue, Olga Babko-Malaya, Ben Snyder) CLSW 2011 NTU, Taipei May 4, 2011

EXAMPLES OF SEMANTIC ANNOTATIONS Predicators and their named arguments � Anaphors and their antecedents

EXAMPLES OF SEMANTIC ANNOTATIONS Predicators and their named arguments � Anaphors and their antecedents � [The protein] inhibits growth in yeast. [It] blocks production… Acronyms and their long forms � [The man]agent painted [the wall]patient. [Platelet-derived growth factor] (known as [pdgf]) impacts … Semantic Typing of entities � [The man]human fired [the gun]firearm

LINGUISTIC PHENOMENA Syntactic Structure � Predicate Argument Structure � Who did what to whom:

LINGUISTIC PHENOMENA Syntactic Structure � Predicate Argument Structure � Who did what to whom: Subject, object, predicate Temporal Structure � Describes grammatical arrangements of words into hierarchical structure Temporal ordering and anchoring of events in a text Emotive and Discourse Structure � How language is used across sentences, and how content is expressed emotionally. Annotated corpora allow us to evaluate and train systems to be able to make these distinctions

MOTIVATION OF ANNOTATION Semantic annotation is critical for robust language understanding � Annotation schemata

MOTIVATION OF ANNOTATION Semantic annotation is critical for robust language understanding � Annotation schemata should focus on a single coherent theme: � Question answering, summarization, inference, reading, … Different linguistic phenomena should be annotated separately over the same corpus The Annotate, Train, and Test Model advances linguistic theory: Theories needs testing to evaluate coverage and predictive force. � Semantic theories are too complex to develop without this model. �

METHODOLOGICAL ASSUMPTION Annotation scheme: � assumes a given feature set Feature set: � encodes

METHODOLOGICAL ASSUMPTION Annotation scheme: � assumes a given feature set Feature set: � encodes specific structural descriptions and properties of the input data Structural descriptions: � theoretically-informed attributes derived from empirical observations over the data Theory Description Features Annotation

LINGUISTIC ANNOTATION SCHEMES Prop. Bank • – Palmer, Gildea, and Kingsbury (2005) Nom. Bank

LINGUISTIC ANNOTATION SCHEMES Prop. Bank • – Palmer, Gildea, and Kingsbury (2005) Nom. Bank • – Meyers, Reeves, Macleod, Szekely, Zielinska, Young, and Grishman (2004) Time. Bank • – Pustejovsky, Littman, Knippen, and Sauri (2005) Opinion Corpus • – Wiebe, Wilson, and Cardie (2005) Penn Discourse Tree. Bank • – Miltsakaki, Prasad, Joshi, and Webber (2004)

PROPBANK • • • Corpus annotated with semantic roles for arguments and adjuncts of

PROPBANK • • • Corpus annotated with semantic roles for arguments and adjuncts of verbs 1 M word Penn Treebank II WSJ corpus. Coarse-grained sense tags, based on grouping of Word. Net senses

PROPOSITION BANK: FROM SENTENCES TO PROPOSITIONS Powell met Zhu Rongji Powell and Zhu Rongji

PROPOSITION BANK: FROM SENTENCES TO PROPOSITIONS Powell met Zhu Rongji Powell and Zhu Rongji met Powell met with Zhu Rongji Powell and Zhu Rongji had a meeting . . . Proposition: meet(Powell, Zhu Rongji) meet(Somebody 1, Somebody 2)

PROPBANK ANNOTATION EXAMPLE [Arg. M-ADV According to reports], [Arg 1 sea trials for [Arg

PROPBANK ANNOTATION EXAMPLE [Arg. M-ADV According to reports], [Arg 1 sea trials for [Arg 1 a patrol boat] [Rel_develop. 02 developed] [Arg 0 by Kazakhstan]] are being [Rel_conduct. 01 conducted] and [Arg 1 the formal launch] is [Rel_plan. 01 planned] [Arg. M-TMP for the beginning of April this year].

PROPOSITION BANK: FROM SENTENCES TO PROPOSITIONS Powell met Zhu Rongji Powell and Zhu Rongji

PROPOSITION BANK: FROM SENTENCES TO PROPOSITIONS Powell met Zhu Rongji Powell and Zhu Rongji met Powell met with Zhu Rongji Powell and Zhu Rongji had a meeting . . . Proposition: meet(Powell, Zhu Rongji) meet(Somebody 1, Somebody 2)

PROPBANK ANNOTATION EXAMPLE [Arg. M-ADV According to reports], [Arg 1 sea trials for [Arg

PROPBANK ANNOTATION EXAMPLE [Arg. M-ADV According to reports], [Arg 1 sea trials for [Arg 1 a patrol boat] [Rel_develop. 02 developed] [Arg 0 by Kazakhstan]] are being [Rel_conduct. 01 conducted] and [Arg 1 the formal launch] is [Rel_plan. 01 planned] [Arg. M-TMP for the beginning of April this year].

WHAT IS A PROPBANK? A Prop. Bank is a corpus annotated with the predicateargument

WHAT IS A PROPBANK? A Prop. Bank is a corpus annotated with the predicateargument structure of the verbs: � English Propbank: www. cis. upenn. edu/~ace 3/’ 04 LDC Kingsbury and Palmer 2002, Palmer, Gildea, Kingsbury, 2005 Wall Street Journal, 1 M words, 120 K+ predicate instances Brown, 14 K predicate instances � Chinese Propbank: www. cis. upenn. edu/~chinese/cpb Xue and Palmer 2003, Xue 2004 Xinhua (250 K words – almost done), Sinorama (250 K words – estimated 2007) Nominalized verbs for English = Nom. Bank/NYU Chinese Nom. Bank?

CAPTURING “NEUTRAL” SEMANTIC ROLES Boyan broke [ Arg 1 the LCD-projector. ] break (agent(Boyan),

CAPTURING “NEUTRAL” SEMANTIC ROLES Boyan broke [ Arg 1 the LCD-projector. ] break (agent(Boyan), patient(LCD-projector)) [Arg 1 The windows] were broken by the hurricane. [Arg 1 The vase] broke into pieces when it toppled over

FRAMES FILE EXAMPLE: GIVE < 4000 FRAMES FOR PROPBANK Roles: Arg 0: giver Arg

FRAMES FILE EXAMPLE: GIVE < 4000 FRAMES FOR PROPBANK Roles: Arg 0: giver Arg 1: thing given Arg 2: entity given to Example: double object The executives gave the chefs a standing ovation. Arg 0: The executives REL: gave Arg 2: the chefs Arg 1: a standing ovation

FRAMES FILE EXAMPLE: GIVE W/ THEMATIC ROLE LABELS Roles: Arg 0: giver Arg 1:

FRAMES FILE EXAMPLE: GIVE W/ THEMATIC ROLE LABELS Roles: Arg 0: giver Arg 1: thing given Arg 2: entity given to Example: double object The executives gave the chefs a standing ovation. Arg 0: Agent The executives REL: gave Arg 2: Recipient the chefs Arg 1: Theme a standing ovation Verb. Net – based on Levin classes

PROPBANK EXERCISE EX. [He]-Arg 1 Theme [will]-MOD [probably]-MOD be [extradited]-rel [to the U. S]-DIR

PROPBANK EXERCISE EX. [He]-Arg 1 Theme [will]-MOD [probably]-MOD be [extradited]-rel [to the U. S]-DIR [for trial under an extradition treaty President Virgilia Barco has revived]PRP. He will probably be extradited to the U. S for trial under [an extradition treaty]-Arg 1 Theme [President Virgilia Barco]-Arg 0 Agent has [revived]-rel.

A CHINESE TREEBANK SENTENCE 国会/Congress 最近/recently 通过/pass 了/ASP 银行法 /banking law “The Congress passed

A CHINESE TREEBANK SENTENCE 国会/Congress 最近/recently 通过/pass 了/ASP 银行法 /banking law “The Congress passed the banking law recently. ” (IP (NP-SBJ (NN 国会/Congress)) (VP (ADV 最近/recently)) (VP (VV 通过/pass) (AS 了/ASP) (NP-OBJ (NN 银行法/banking law)))))

THE SAME SENTENCE, PROPBANKED (IP (NP-SBJ arg 0 (NN 国会)) (VP arg. M (ADVP

THE SAME SENTENCE, PROPBANKED (IP (NP-SBJ arg 0 (NN 国会)) (VP arg. M (ADVP (ADV 最近)) (VP f 2 (VV 通过) (AS 了) 通过(f 2) (pass) arg 1 (NP-OBJ (NN 银行法 ))))) arg 0 arg. M arg 1 国会 最近 银行法 (law) (congress)

ANNOTATION PROCEDURE PTB II – Extract all sentences of a verb Create Frame File

ANNOTATION PROCEDURE PTB II – Extract all sentences of a verb Create Frame File for that verb Paul Kingsbury (3400+ lemmas, 4700 framesets, 120 K predicates) 1 st pass: Automatic tagging Joseph Rosenzweig 2 nd pass: Double blind hand correction by verb Inter-annotator agreement 84% (87% Arg#’s) 3 rd pass: Adjudication Olga Babko-Malaya 4 th pass: Train automatic semantic role labellers Dan Gildea, Sameer Pradhan, Nianwen Xue, Szuting Yi, …. Co. NLL-04 shared task, 2004, 2005, ….

WORD SENSES IN PROPBANK Orders to ignore word sense not feasible for 700+ verbs

WORD SENSES IN PROPBANK Orders to ignore word sense not feasible for 700+ verbs � � Mary left the room Mary left her daughter-in-law her pearls in her will Frameset leave. 01 "move away from": Arg 0: entity leaving Arg 1: place left Frameset leave. 02 "give": Arg 0: giver Arg 1: thing given Arg 2: beneficiary How do these relate to traditional word senses in Word. Net?

PROPBANK II – ENGLISH/CHINESE (100 K) We still need relations between events and entities:

PROPBANK II – ENGLISH/CHINESE (100 K) We still need relations between events and entities: Event ID’s with event coreference Selective sense tagging � Tagging nominalizations w/ Word. Net sense � Grouped WN senses - selected verbs and nouns Nominal Coreference � not names Clausal Discourse connectives – selected subset Level of representation that reconciles many surface differences between the languages

EVENT IDS – PARALLEL PROP II (1) Aspectual verbs do not receive event IDs:

EVENT IDS – PARALLEL PROP II (1) Aspectual verbs do not receive event IDs: � 今年/this year 中国/China �� /continue �� /play 其/it 在/at 支持/support 外商/foreign business 投� /investment 企� /enterprise 方面/aspect 的/DE 主/main 渠道/channel 作用/role “This year, the Bank of China will continue to play the main role in supporting foreign-invested businesses. ”

EVENT IDS – PARALLEL PROP II (2) Nominalized verbs do: � He will probably

EVENT IDS – PARALLEL PROP II (2) Nominalized verbs do: � He will probably be extradited to the US for trial. done as part of sense-tagging (all 7 WN senses for “trial” are events. ) � 随着/with 中国/China �� /economy 的/DE 不断 /continued �展 /development… “With the continued development of China’s economy…” The same events may be described by verbs in English and nouns in Chinese, or vice versa. Event IDs help to abstract away from POS tag

EVENT REFERENCE – PARALLEL PROP II Pronouns (overt or covert) that refer to events:

EVENT REFERENCE – PARALLEL PROP II Pronouns (overt or covert) that refer to events: [This] is gonna be a word of mouth kind of thing. �些 /these 成果/achivements 被/BEI 企� /enterprise 用 /apply (e 15) 到/to 生� /production 上/on 点石成金/spin gold from straw, *pro*-e 15 大大/greatly 提高/improve 了 /le 中国/China � /nickel � /industry 的/DE 生� /production 水平/level 。 “These achievements have been applied (e 15) to production by enterprises to spin gold from straw, which-e 15 greatly improved the production level of China’s nickel industry. ” Prerequisites: � pronoun classification � free trace annotation

CHINESE PB II: SENSE TAGGING Much lower polysemy than English � Avg of 3.

CHINESE PB II: SENSE TAGGING Much lower polysemy than English � Avg of 3. 5 (Chinese) vs. 16. 7 (English) Dang, Chia, Chiou, Palmer, COLING-02 � More than 2 Framesets 62/4865 (250 K) Ch vs. 294/3635 (1 M) English Mapping Grouped English senses to Chinese (English tagging - 93 verbs/168 nouns, 5000+ instances) Selected 12 polysemous English words (7 verbs/5 nouns) � For 9 (6 verbs/3 nouns), grouped English senses map to unique Chinese translation sets (synonyms) �

MAPPING OF GROUPED SENSE TAGS TO CHINESE increase 提高 / ti 2 gao 1

MAPPING OF GROUPED SENSE TAGS TO CHINESE increase 提高 / ti 2 gao 1 Collect, levy 募集 / mu 4 ji 2 筹措 / chou 2 cuo 4 筹. . . / chou 2… raise – translations by group lift, elevate, orient upwards 仰 / yang 3 invoke, elicit, set off 提 / ti 4

DISCOURSE CONNECTIVES: THE PENN DISCOURSE TREEBANK WSJ corpus (~1 M words, ~2400 texts) http:

DISCOURSE CONNECTIVES: THE PENN DISCOURSE TREEBANK WSJ corpus (~1 M words, ~2400 texts) http: //www. cis. upenn. edu/~pdtb Miltsakaki, Prasad, Joshi and Webber, LREC-04, NAACL-04 Frontiers Prasad, Miltsakaki, Joshi and Webber ACL-04 Discourse Annotation Chinese: 10 explicit discourse connectives that include subordination conjunctions, coordinate conjunctions, and discourse adverbials. Argument determination, sense disambiguation [arg 1 学校/school 不/not 教/teach 理� /finance management], [conn �果 /as a result] [arg 2 �章 /newspaper 上/on 的/DE 各/all 种/kind �� /column 就/then 成� /become 信息/information 的/DE 主要/main 来源 /source]。 “The school does not teach finance management. As a result, the different kinds of columns become the main source of information. ”

MAPPING OF GROUPED SENSE TAGS TO CHINESE � Zhejiang|浙江zhe 4 jiang 1 will|将jiang 1

MAPPING OF GROUPED SENSE TAGS TO CHINESE � Zhejiang|浙江zhe 4 jiang 1 will|将jiang 1 raise|提高ti 2 gao 1 the level|水平shui 3 ping 2 of|的de opening up|开放kai 1 fang 4 to|对 dui 4 the outside world|外wai 4. (浙江将提高对外开放的水平。) � I|我wo 3 raised|仰yang 3 my|我的wo 3 de head|头tou 2 in expectation|期望qi 1 wang 4. (我仰头望去。) � …, raising|筹措chou 2 cuo 4 funds|资金zi 1 jin 1 of|的de 15 billion|150亿yi 1 ban 3 wu 3 shi 2 yi 4 yuan|元yuan 2 (…筹措资金 150 亿元。) � The meeting|会议hui 4 yi 4 passed|通过tong 1 guo 4 the “decision regarding motions”|议案yi 4 an 4 raised|提ti 4 by 32 NPC|人大 ren 2 da 4 representatives|代表dai 4 biao 3 (会议通过了32名人大代 表所提的议案。)

NOMBANK • • Provides argument structure for 5000 common noun lemmas from the Penn

NOMBANK • • Provides argument structure for 5000 common noun lemmas from the Penn Treebank II corpus. Borrows heavily from Prop. Bank where possible (for example for nominalizations)

NOMBANK EXAMPLES Verb-Related � Adjective Related � Powell’s/ARG 0 meeting with Zhu Rongji/ARG 1

NOMBANK EXAMPLES Verb-Related � Adjective Related � Powell’s/ARG 0 meeting with Zhu Rongji/ARG 1 The absence of patent lawyers/ARG 1 in the court/ARG 2 Nominals (16 classes) Her/ARG 1 husband/ARG 0 � An Oct. 1/ARG 2 date for the attack/ARG 1 �

NOMBANK ANNOTATION EXAMPLE According to [Rel_report. 01 reports], [Arg 1 sea [Rel_trial. 01 trials]

NOMBANK ANNOTATION EXAMPLE According to [Rel_report. 01 reports], [Arg 1 sea [Rel_trial. 01 trials] [Arg 1 for [Arg 1 -CF_launch. 01 a patrol boat] developed by Kazakhstan] are being conducted and the [Arg. M-MNR formal] [Rel_launch. 01 launch] is planned for the [[REL_beginning. 01 beginning] [ARG 1 of April this year]].

OPINION ANNOTATION I think people are happy because Chavez has fallen. direct subjective span:

OPINION ANNOTATION I think people are happy because Chavez has fallen. direct subjective span: think source: <writer, I> attitude: attitude span: think type: positive arguing intensity: medium target: target span: people are happy because Chavez has fallen direct subjective span: are happy source: <writer, I, People> attitude: attitude span: are happy type: pos sentiment intensity: medium target: target span: Chavez has fallen inferred attitude span: are happy because Chavez has fallen type: neg sentiment intensity: medium target: target span: Chavez

MOTIVATING EXAMPLE “I think people are happy because Chavez has fallen. But there’s also

MOTIVATING EXAMPLE “I think people are happy because Chavez has fallen. But there’s also a feeling of uncertainty about how the country’s obvious problems are going to be solved, ” said Ms. Ledesma. 7/27/2004 AAAI 2004 33

MOTIVATING EXAMPLE medium strength Though some of them did not conceal their criticisms of

MOTIVATING EXAMPLE medium strength Though some of them did not conceal their criticisms of Hugo Chavez, the member countries of the Organization of American States condemned the coup and recognized the legitimacy of the elected president. high strength low strength

PRIVATE STATES AND SUBJECTIVE EXPRESSIONS Private state: covering term for opinions, emotions, sentiments, attitudes,

PRIVATE STATES AND SUBJECTIVE EXPRESSIONS Private state: covering term for opinions, emotions, sentiments, attitudes, speculations, etc. (Quirk et al. , 1985) Subjective Expressions: words and phrases that express private states (Banfield, 1982) “The US fears a spill-over, ” said Xirao-Nima. “The report is full of absurdities, ” he complained.

CORPUS OF OPINION ANNOTATIONS Multi-perspective Question Answering (MPQA) Corpus � Sponsored by NRRC ARDA

CORPUS OF OPINION ANNOTATIONS Multi-perspective Question Answering (MPQA) Corpus � Sponsored by NRRC ARDA � Released November, 2003 � http: //nrrc. mitre. org/NRRC/publications. htm Detailed expression-level annotations of private states: strength See Wilson and Wiebe (SIGdial 2003) Freely Available

PENN DISCOURSE TREEBANK (PDTB) • • Annotate discourse connectives and their arguments Discourse connectives

PENN DISCOURSE TREEBANK (PDTB) • • Annotate discourse connectives and their arguments Discourse connectives take clauses as their arguments and express relations between clauses – • i. e. , relations between propositions, events, situations Discourse connectives such as - and, or, but, because, since, while, when, however, instead, although, also, for example, then, so that, insofar as, nonetheless • Subordinate conjunctions, Coordinate conjunctions, Adverbial connectives, Implicit connectives • • Because [Arg 2 he was sick], [Arg 1 John left early] Since [Arg 2 the store is closed], [Arg 1 we’ll go home].

THE PROBLEM Connective Arg 2 After adjusting for inflation, the Commerce Department said spending

THE PROBLEM Connective Arg 2 After adjusting for inflation, the Commerce Department said spending didn’t change in September. Arg 1 After adjusting for inflation, the Commerce Department said spending didn’t change in September. Given a discourse connective, identify the heads of its two arguments

IDENTIFYING ARGUMENTS IN PDTB Task Identify lexicalized relations in Penn Discourse Tree. Bank (PDTB)

IDENTIFYING ARGUMENTS IN PDTB Task Identify lexicalized relations in Penn Discourse Tree. Bank (PDTB) � Identify head-words of arguments � Don’t identify relation type or non-lexicalized relations � Approach � Rank Arg 1 & Arg 2 candidate arguments separately � Re-rank top N argument pairs Apply Max. Ent statistical ranker Model both argument candidates jointly Re-ranking reduces error 5 -11% Main Results: 74% accuracy at identifying both arguments correctly for a connective � Using gold-standard Tree. Bank parses

PDTB EXAMPLES Coordinator Choose 203 business executives, including, perhaps, someone from your own staff,

PDTB EXAMPLES Coordinator Choose 203 business executives, including, perhaps, someone from your own staff, and put them out on the streets, to be deprived for one month of their homes, families and income. Subordinator Drug makers shouldn’t be able to duck liability because people couldn’t identify precisely which identical drug was used. France’s second-largest government-owned insurance company, Assurances Generales de France, has been building its own Naviation Mixte stake, currently thought to be between 8% and 10%. Analysts said they don’t think it is contemplating a takeover, however, and its officials couldn’t be reached. Discourse Adverbial

MOTIVATION FOR TIME AND EVENT MARKUP Natural language is filled with references to past

MOTIVATION FOR TIME AND EVENT MARKUP Natural language is filled with references to past and future events, as well as planned activities and goals; Without a robust ability to identify and temporally situate events of interest from language, the real importance of the information can be missed; A Robust Annotation standard can help leverage this information from natural language text.

TEMPORAL AWARENESS IN REAL TEXT The bridge collapsed during the storm but after traffic

TEMPORAL AWARENESS IN REAL TEXT The bridge collapsed during the storm but after traffic was rerouted to the Bay Bridge. President Roosevelt died in April 1945 before � the war ended. (event happened) � he dropped the bomb. (event didn’t happen) The CEO plans to retire next month. Last week Bill was running the marathon when he twisted his ankle. Someone had tripped him. He fell and didn't finish the race.

CURRENT TIME ANALYSIS TECHNOLOGY Document Time Linking � Find the document creation time and

CURRENT TIME ANALYSIS TECHNOLOGY Document Time Linking � Find the document creation time and link that to all events in the text; Local Time Stamping � find an event and a “local temporal expression”, and link it to that time;

DOCUMENT TIME STAMPING April 25, 2010 President Obama paid tribute Sunday to 29 workers

DOCUMENT TIME STAMPING April 25, 2010 President Obama paid tribute Sunday to 29 workers killed in an explosion at a West Virginia coal mine earlier this month, saying they died "in pursuit of the American dream. " The blast at the Upper Big Branch Mine was the worst U. S. mine disaster in nearly 40 years. Obama ordered a review earlier this month and blamed mine officials for lax regulation.

DOCUMENT TIME STAMPING: April 25, 2010 President Obama paid tribute Sunday to 29 workers

DOCUMENT TIME STAMPING: April 25, 2010 President Obama paid tribute Sunday to 29 workers killed in an explosion at a West Virginia coal mine earlier this month, saying they died "in pursuit of the American dream. " The blast at the Upper Big Branch Mine was the worst U. S. mine disaster in nearly 40 years. Obama ordered a review earlier this month and blamed mine officials for lax regulation.

DOCUMENT TIME STAMPING: FOR REAL April 25, 2010 President Obama paid tribute Sunday to

DOCUMENT TIME STAMPING: FOR REAL April 25, 2010 President Obama paid tribute Sunday to 29 workers killed in an explosion at a West Virginia coal mine earlier this month, saying they died "in pursuit of the American dream. " The blast at the Upper Big Branch Mine was the worst U. S. mine disaster in nearly 40 years. Obama ordered a review earlier this month and blamed mine officials for lax regulation.

TIME STAMPING: THE GOOD, BAD, … ✓ ☺Set up a meeting on Tuesday with

TIME STAMPING: THE GOOD, BAD, … ✓ ☺Set up a meeting on Tuesday with EMC. ✓ ☺Franklin arrives tomorrow from London. ✗ ☹ Franklin arrives on the afternoon flight from London tomorrow. ✗ ☹ ☹ Most people drive today while talking on the phone.

TEMPORAL AWARENESS CHALLENGE Identification of all important events in a text Actual temporal ordering

TEMPORAL AWARENESS CHALLENGE Identification of all important events in a text Actual temporal ordering and time anchoring of these events to temporal expressions.

ISO-TIMEML ENABLES TEMPORAL PARSING A new generation of language analysis tools that are able

ISO-TIMEML ENABLES TEMPORAL PARSING A new generation of language analysis tools that are able to temporally organize events in terms of their ordering and time of occurrence These tools can be integrated with visualization, summarization, question answering, and link analysis systems to help analyze large event-rich information spaces.

ISO-TIMEML PROVIDES ELEMENTS TO: Find all events and times in newswire text Link events

ISO-TIMEML PROVIDES ELEMENTS TO: Find all events and times in newswire text Link events to the document time and to local times Order event relative to other events Ensure consistency of the temporal relations

TEMPORAL PARSING TECHNOLOGIES Build temporal representations of events in document collections; Track people and

TEMPORAL PARSING TECHNOLOGIES Build temporal representations of events in document collections; Track people and the events they participated in; Answer questions about when events occur.

APPLICATIONS IMPACTED Health Care, Bioinformatics, Insurance Object Tracking Search and Categorization Trend Analysis and

APPLICATIONS IMPACTED Health Care, Bioinformatics, Insurance Object Tracking Search and Categorization Trend Analysis and Prediction

TEMPORAL AWARENESS Take your 1 st dose of levaquin in the morning before any

TEMPORAL AWARENESS Take your 1 st dose of levaquin in the morning before any food, 2 nd dose before sleep. dose 1 eat dose 2 … sleep

TEMPORAL AWARENESS No food or drink after midnight before surgery, until you are in

TEMPORAL AWARENESS No food or drink after midnight before surgery, until you are in recovery. 12: 00 am surgery ¬food &¬drink food & drink recovery

DIFFERENT NOTIONS OF EVENTS Topic: “well-defined subject” for searching � Template: structure with slots

DIFFERENT NOTIONS OF EVENTS Topic: “well-defined subject” for searching � Template: structure with slots for participant named entities � document- or collection-level document-level Mention: linguistic expression that expresses an underlying event � phrase-level (verb/noun)

EVENTS AS TEMPLATES Wall Street Journal, 06/15/88 MAXICARE HEALTH PLANS INC and UNIVERSAL HEALTH

EVENTS AS TEMPLATES Wall Street Journal, 06/15/88 MAXICARE HEALTH PLANS INC and UNIVERSAL HEALTH SERVICES INC have dissolved a joint venture which provided health services. Systems can fill such templates at ~ 60% accuracy from news (MUC evals)

ACE EVENT TYPES

ACE EVENT TYPES

ACE Event Roles

ACE Event Roles

EVENTS IN TIMEML Mention: linguistic expression that expresses an underlying event � Since they

EVENTS IN TIMEML Mention: linguistic expression that expresses an underlying event � Since they correspond to surface mentions, easier to annotate and recognize � Accuracy is > 88% (ARDA AQUAINT (TARSQI)) Like templates � Phrase-level (verb/noun) they are linked to times Unlike templates � the times are resolved � � 87% accuracy in time resolution (TERN evals: timex 2. mitre. org) the links involve temporal relations the events are temporally ordered the links also involve other logical relations (subordinating and aspectual)

FEATURES OF ISO-TIMEML Identifies temporal expressions; � � � Identifies signals determining interpretation of

FEATURES OF ISO-TIMEML Identifies temporal expressions; � � � Identifies signals determining interpretation of temporal expressions; � � Temporal Prepositions: for, during, on, at; Temporal Connectives: before, after, while. Identifies event expressions; � � � Dates, times Temporal Functions: three years ago Anchors to events and other temporal expressions: three years after the Gulf War tensed verbs; has left, was captured, will resign; stative adjectives; sunken, stalled, on board; event nominals; merger, Military Operation, Gulf War; Creates dependencies between events and times: � � � Anchoring; John left on Monday. Orderings; The party happened after midnight. Embedding; John said Mary left.

ISO-TIMEML TAGS <TIMEX 3> � Used to mark up explicit temporal expressions, such as

ISO-TIMEML TAGS <TIMEX 3> � Used to mark up explicit temporal expressions, such as times, dates, durations, etc. It is modeled on the TIDES TIMEX 2 tag. <EVENT> � Used to annotate those elements in a text that mark the semantic events described by it. Events are typically verbs, although event nominals, such as "crash" in ". . . killed by the crash", are also annotated as events. <TLINK> � One of the three Time. ML link tags. Link tags encode the various relations that exist between the temporal elements of a document. A TLINK is a temporal link. It represents the relation between two temporal elements. <SLINK> � A subordination link that is used for contexts involving modality, evidentials, and factives. An SLINK is used in cases where an event instance subordinates another event instance type. <ALINK> � An aspectual link, it indicates an aspectual connection between two events. In some ways, it is like a cross between TLINK and SLINK in that it indicates both a relation between two temporal elements, as well as aspectual subordination. <ARGLINK> � A link establishing a relationship between an event and each of its participants. ARGLINK uses the entity ID and binds it to the event.

TIMEML: ANNOTATION OF TEMPORAL ENTITIES Temporal expressions: <TIMEX 3> � � Times: 3 o’clock,

TIMEML: ANNOTATION OF TEMPORAL ENTITIES Temporal expressions: <TIMEX 3> � � Times: 3 o’clock, mid-morning. Dates: Fully Specified: June 11, 1989; Summer, 2002. Underspecified: Monday, next month, two days ago. � Durations: three months, two years. � Sets: every month, each Tuesday. Event expressions: <EVENT> � Expressions denoting events that participate in the narrative of a given document, and which can be temporally ordered. � Event-related grammatical features: Tense: past, present, past, etc. Aspect: progressive, perfective-progressive. Polarity: positive, negative. Modality: would, could, may, etc. Class: occurrence, state, aspectual, intensional, etc.

TIMEML: ANNOTATION OF TEMPORAL RELATIONS Temporal links: <TLINK> Anchoring of Events to Times �

TIMEML: ANNOTATION OF TEMPORAL RELATIONS Temporal links: <TLINK> Anchoring of Events to Times � Ordering of Events � 13 temporal relations (based on Allen’s relations), among which: � • • Simultaneous Before (e. g. , For most of the murders, suspects have already been arrested) After Immediately before (e. g. , All passengers died when the plane crashed into the mountain) Immediately after. Including (e. g. , John arrived in Boston last Thursday) Etc. Aspectual links: <ALINK> Phases of an event Initiation: John started to read. Culmination: John finished assembling the table. Termination: John stopped talking. Continuation: John kept talking.

TIMEML: ANNOTATION OF TEMPORAL RELATIONS Subordinating links <SLINK> � � Events that syntactically subordinate

TIMEML: ANNOTATION OF TEMPORAL RELATIONS Subordinating links <SLINK> � � Events that syntactically subordinate other events Providing information about the factual nature of the embedded event: Factive: The embedded event is presupposed or entailed as factual. John forgot that he was in Boston last year. Mary regrets that she didn't marry John. Counterfactive: The embedded event is presupposed as non-factual: John forgot to buy some wine. John prevented the divorce. Evidential: Introduced by REPORTING or PERCEPTION: John said he bought some wine. Mary saw John carrying only beer. Negative evidential: Introduced by REPORTING events conveying negative polarity: John denied he bought only beer. Modal: Expressing different degrees of uncertainty, possibility, thought, etc. Analysts also suspect suppliers have fallen victim to their own success.

EXAMPLE: TEMPORAL EXPRESSIONS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S.

EXAMPLE: TEMPORAL EXPRESSIONS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and Arab troops at the Saudi border, today sought peace on another front by promising to withdraw from Iranian territory and release soldiers captured during the Iran-Iraq war. Also today, King Hussein of Jordan arrived in Washington seeking to mediate the Persian Gulf crisis. President Bush on Tuesday said the United States may extend its naval quarantine to Jordan's Red Sea port of Aqaba to shut off Iraq's last unhindered trade route. In another mediation effort, the Soviet Union said today it had sent an envoy to the Middle East on a series of stops to include Baghdad. Soviet officials also said Soviet women, children and invalids would be allowed to leave Iraq.

EXAMPLE: EVENTS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and

EXAMPLE: EVENTS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and Arab troops at the Saudi border, today sought peace on another front by promising to withdraw from Iranian territory and release soldiers captured during the Iran-Iraq war. Also today, King Hussein of Jordan arrived in Washington seeking to mediate the Persian Gulf crisis. President Bush on Tuesday said the United States may extend its naval quarantine to Jordan's Red Sea port of Aqaba to shut off Iraq's last unhindered trade route. In another mediation effort, the Soviet Union said today it had sent an envoy to the Middle East on a series of stops to include Baghdad. Soviet officials also said Soviet women, children and invalids would be allowed to leave Iraq.

EXAMPLE: TLINKS, ANCHORING EVENT TO TIMEX AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein,

EXAMPLE: TLINKS, ANCHORING EVENT TO TIMEX AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and Arab troops at the Saudi border, today sought peace on another front by promising to withdraw from Iranian territory and release soldiers captured during the Iran-Iraq war. Also today, King Hussein of Jordan arrived in Washington seeking to mediate the Persian Gulf crisis. President Bush on Tuesday said the United States may extend its naval quarantine to Jordan's Red Sea port of Aqaba to shut off Iraq's last unhindered trade route. In another mediation effort, the Soviet Union said today it had sent an envoy to the Middle East on a series of stops to include Baghdad. Soviet officials also said Soviet women, children and invalids would be allowed to leave Iraq.

EXAMPLE: TLINKS, ORDERING EVENTS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U.

EXAMPLE: TLINKS, ORDERING EVENTS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and Arab troops at the Saudi border, today sought peace on another front by promising to withdraw from Iranian territory and release soldiers captured during the Iran-Iraq war. Also today, King Hussein of Jordan arrived in Washington seeking to mediate the Persian Gulf crisis. President Bush on Tuesday said the United States may extend its naval quarantine to Jordan's Red Sea port of Aqaba to shut off Iraq's last unhindered trade route. Past < Tuesday < Today < Indef Future ______________________________________ war(I, I) say(Bush, S) arrive(H, DC) withdraw(Saddam) captured(sold) seek(Saddam, peace) release(Saddam, soldiers) extend(US, quarantine) shut_off(US, trade_route)

EXAMPLE: ALINKS, PHASES OF EVENTS President Bush today denounced Saddam's ``ruinous policies of war,

EXAMPLE: ALINKS, PHASES OF EVENTS President Bush today denounced Saddam's ``ruinous policies of war, '' and said the United States is ``striking a blow for the principle that might does not make right. '' In a speech delivered at the Pentagon, Bush seemed to suggest that American forces could be in the gulf region for some time. ``No one should doubt our staying power or determination, '' he said. The U. S. military buildup in Saudi Arabia continued at fever pace, with Syrian troops now part of a multinational force camped out in the desert to guard the Saudi kingdom from any new thrust by Iraq. In a letter to President Hashemi Rafsanjani of Iran, read by a broadcaster over Baghdad radio, Saddam said he will begin withdrawing troops from Iranian territory a week from tomorrow and release Iranian prisoners of war.

EXAMPLE: SLINKS, FACTUAL NATURE OFE VENTS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein,

EXAMPLE: SLINKS, FACTUAL NATURE OFE VENTS AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and Arab troops at the Saudi border, today sought peace on another front by promising to withdraw from Iranian territory and release soldiers captured during the Iran-Iraq war. Also today, King Hussein of Jordan arrived in Washington seeking to mediate the Persian Gulf crisis. President Bush on Tuesday said the United States may extend its naval quarantine to Jordan's Red Sea port of Aqaba to shut off Iraq's last trade route. In another mediation effort, the Soviet Union said today it had sent an envoy to the Middle East on a series of stops to include Baghdad. Soviet officials also said Soviet women, children and invalids would be allowed to leave Iraq.

EXAMPLE: SLINKS, REPORTED SPEECH AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U.

EXAMPLE: SLINKS, REPORTED SPEECH AP-NR-08 -15 -90 1337 EDT Iraq's Saddam Hussein, facing U. S. and Arab troops at the Saudi border, today sought peace on another front by promising to withdraw from Iranian territory and release soldiers captured during the Iran-Iraq war. Also today, King Hussein of Jordan arrived in Washington seeking to mediate the Persian Gulf crisis. President Bush on Tuesday said the United States may extend its naval quarantine to Jordan's Red Sea port of Aqaba to shut off Iraq's last unhindered trade route. In another mediation effort, the Soviet Union said today it had sent an envoy to the Middle East on a series of stops to include Baghdad. Soviet officials also said Soviet women, children and invalids would be allowed to leave Iraq.

MODELING EVENTS RELATIVE TO TIME: ORDER: � The position of the interval relative to

MODELING EVENTS RELATIVE TO TIME: ORDER: � The position of the interval relative to others : MEASURE: � The size of the interval; QUANTITY: � The number of intervals.

ORDER John taught on Tuesday. John taught before Mary arrived.

ORDER John taught on Tuesday. John taught before Mary arrived.

MEASURE John taught for three hours on Tuesday. Introduce MLINK: <EVENT id="e 1" pred="TEACH"/>

MEASURE John taught for three hours on Tuesday. Introduce MLINK: <EVENT id="e 1" pred="TEACH"/> <TIMEX 3 id="t 2" type="DURATION" value="P 3 H"/> <MLINK event. ID="e 1" related. To. Time="t 2" />

QUANTITY John taught every Monday in November.

QUANTITY John taught every Monday in November.

SUMMARY OF ISO-TIMEML Enhances our ability to annotate temporal and event expressions in multiple

SUMMARY OF ISO-TIMEML Enhances our ability to annotate temporal and event expressions in multiple languages Has an explicit semantics associated with the abstract syntactic specification Is already being tested against Sem. Eval standards competetions. Integrated into the TTK (TARSQI Toolkit) at Brandeis

THANK YOU! timeml. org

THANK YOU! timeml. org