NLP Introduction to NLP FiniteState Automata 22 slides

NLP

Introduction to NLP Finite-State Automata (2/2) (slides from Jurafsky and Martin)

NFSA - Figure 2. 17

Epsilon transitions - Figure 2. 18

Three tenets of NFSA • Backup • Lookahead • Parallelism

Figure 2. 19

LIFO - Figure 2. 20

FIFO - Figure 2. 21

Regular languages

Closures

Notes about regular languages NFSA and DFSA are equivalent. How? Converting regexps into automata …

Base case - Figure 2. 22

Figure 2. 23

Figure 2. 24

Figure 2. 25

Assignment 2. 8 - Figure 2. 26 Write a regular language for this NFSA.

The FSM toolkit and friends • Developed at AT&T Research – Riley, Pereira, Mohri, Sproat • 4 useful parts – FSM, Lextools, GRM, Dot (separate)

Introduction to NLP Finite State Transducers

FSA for adjectives • • Big, bigger, biggest Cool, cooler, coolest, coolly Red, redder, reddest Clear, clearest, clearly, unclearly Happy, happier, happiest, happily Unhappy, unhappier, unhappiest, unhappily What about: unbig, redly, and realest? Irregular plurals

Notes • Is a string a legitimate word or not? • Two-level morphology: lexical level + surface level (Koskenniemi 83) • Finite-state transducers (FST) – used for regular relations • Inversion and composition of FST • Transducers for transliteration

Exercise • FST for English plurals • Weights?

Transliteration アメフト アイスクリーム アイドル アパート バイク バリアフリー コンピューター デスク ラマ エレベーター エスカレーター フライドポテト グラス ハッピーエンド ハンカチ ホットケーキ カシューナッツ コーヒー クラブ キーボード キャンペーン キャップ パソコン パーソナルコンピューター ペーパーテスト レジュメ レストラン リモコン サラダ タバコ テレビゲーム ゼミナール amefuto aisu kurīmu aidoru apāto baiku bariafurī konpyūtā desuku dorama erebētā esukarētā furaidopoteto gurasu happīendo hankachi hottokēki kashū nattsu kōhī kurabu kībōdo kyanpēn kyappu pāsokon pāsonaru konpyūtā pēpātesuto rejume resutoran rimokon sarada tabako terebigēmu zemināru Ame(rican) Foot(ball) ice cream idol apartment bike barrier free computer desk (at a news agency, e. g. , the Sports desk) drama (on TV) elevator escalator fried potato (French fries) glass (for drinking) happy end(ing) handkerchief hotcake (pancake) cashew nut coffee club keyboard campaign cap perso(nal) com(puter) personal computer paper (written) test resume restaurant remo(te) con(trol) salad tobacco television (video) game seminar

Conjugation Present je vais I go tu vas you go il va he goes nous allons we go vous allez you go ils vont they go Past je suis allé I went tu es allé you went il est allé he went nous sommes allés we went vous êtes allés you went ils sont allés they went Imperfect j'allais I used to go tu allais you used to go il allait he used to go nous allions we used to go vous alliez you used to go ils allaient they used to go Conditional j'irais I would go tu irais you would go il irait he would go nous irions we would go vous iriez you would go ils iraient they would go Future j'irai I will go tu iras you will go il ira he will go nous irons we will go vous irez you will go ils iront they will go Subjunctive que j'aille that I go que tu ailles that you go qu'il aille that he go que nous allions that we go que vous alliez that you go qu'ils aillent that they go

NLP