1. Constraint Based Hindi Parser LTRC, IIIT Hyderabad

2. Introduction Broad coverage parser Very crucial IL-IL MT systems, IE, co-reference resolution, etc.

3. Why Dependency ? Phrase Structures Intrinsically presumes order Context Free Grammar (CFG) not well-suited for free-word order languages (Shieber, 1985) Particularly ill suited to Indian Languages Dependency Structures Gives flexibility Common structures With appropriate labels, closer to Semantics

4. Computational Paninian Grammar (CPG) Based on Panini’s Grammar (500 BC) Inspired by Inflectionally rich language (Sanskrit) A dependency based analysis

5. Computational Paninian Grammar (The Basic Framework) Treats a sentence as a set of modifier- modified relations Sentence has a primary modified or the root (which is generally a verb) Gives us the framework to identify these relations Relations between noun constituent and verb called ‘karaka’ karakas are syntactico-semantic in nature Syntactic cues help us in identifying the karakas

6. karta – karma karaka The boy opened the lock k1 – karta k2 – karma karta, karma usually correspond to agent, theme But not always karakas are direct participants in the activity denoted by the verb open boylock k1k2

7. Basic karaka relations karta – agent/doer/force Relation label – k1 karma – object/patient Relation label – k2 karana – instrument Relation label – k3 sampradaan – beneficiary Relation label – k4 apaadaan – source Relation label – k5 adhikarana – location in place/time/other Relation label – k7p/k7t/k7 For complete list of dependency relations: (Begum et al., 2008)

8. Basic karaka relations raama phala khaataa hai ‘Ram eats fruit’

9. Basic karaka relations raama chaaku se saiv kaatataa hai ‘Ram cuts the apple with knife’

10. Basic karaka relations raama ne mohana ko pustaka dii ‘Ram gave a book to Mohan’

11. Why Paninian Labels Other choices for labels could be Grammatical relations Subject, Object, etc. Behavioral tests (Mohanan, 1994) Thematic roles Agent, patient, etc. No concrete cues Difficult to extract them automatically Karakas can be computationally exploited Syntactically grounded, Semantically loaded Gives a level of interface

12. Levels of Language Analysis Morphological analysis (Morph Info.) Analysis in local context (POS tagging) Sentence analysis (Chunking, Parsing) Semantic analysis (Word sense disambiguation, etc.) Discourse processing (Anaphora resolution, Informational Structure, etc.)

13. Example rAma ne mohana ko puswaka xI |

14. Example – Parsed Output xI ‘give’ puswaka ‘book’ mohanarAma k2 k4 k1

15. Parser Two stage strategy Appropriate constraints formed Stage I (Intra-clausal relations) Dependency relations marked Relations such as k1, k2, k3, etc. for each verb Stage II (Inter-clausal relations & conjunct relations) Conjuncts, relative clauses, kriya mula, etc

16. Demand Frame for Verb A demand frame or karaka frame for a verb indicates the demands the verb makes It depends on the verb and its tense, aspect and modality (TAM) label. A mapping is specified between karaka relations and vibhaktis (post-positions, suffix).

17. Karaka Frame It specifies what karakas are mandatory or optional for the verb and what vibhaktis (post- positions) they take respectively Each verb belongs to a specific verb class Each class has a basic karaka frame Each TAM specifies a transformation rule

18. Example rAma mohana ko puswaka xewA hE | xewA hE ‘give is’ puswaka ‘book’ mohanarAma k2 k4 k1 Parsed Dependency Tree

19. Transformations Based on the TAM of the verb rAma ne mohana ko KilOnA xiyA | rAma ko mohana ko KilOnA xenA padZA | Appropriate transformation applied

20. Example rAma ne mohana ko puswaka xI |

21. Karaka Frame – xe (give)

22. Transformation Rule – yA (TAM)

23. Karaka Frame rAma ne mohana ko KilOnA xiyA | yA TAM ---------------------------------------------------------------------------------------- arc-label necessity vibhakti lextype src-pos arc-dir ---------------------------------------------------------------------------------------- k1 m ne n l c k2 m 0|ko n l c k3 d se n l c k4 d ko n l c ---------------------------------------------------------------------------------------- Transformed frame for xe after applying the yA trasformation 0  ne

24. Parsed Output xI ‘give’ puswaka ‘book’ mohanarAma k2 k4 k1

25. Other frames Adjectives

26. Steps in Parsing Morph, POS tagging, Chunking SENTENCE Identify Demand Groups Load Frames & Transform Find Candidates Apply Constraints & Solve Final Parse

27. Example: rAma ne mohana ko KilOnA xiyA |

28. Identify the demand group, Load and Transform DF xiyA Only verb Transformed frame Use ‘yA’ TAM info. ---------------------------------------------------------------------------------------- arc-label necessity vibhakti lextype src-pos arc-dir ---------------------------------------------------------------------------------------- k1 m ne n l c k2 m 0|ko n l c k3 d se n l c k4 d ko n l c ----------------------------------------------------------------------------------------

29. Candidates rAma ne mohana ko KilOnA xiyA _ROOT_ | k1 k2 k4 k2 main

30. Constraints C1: For each of the mandatory demands in a demand frame for each demand group, there should be exactly one outgoing edge labeled by the demand from the demand group. C2: For each of the optional demands in a demand frame for each demand group, there should be at most one outgoing edge labeled by the demand from the demand group. C3: There should be exactly one incoming arc into each source group.

31. Constraints A parse of a sentence is obtained by satisfying all the above constraints Ambiguous sentences have multiple parses Ill formed sentences have no parse.

32. Parse - I rAma ne mohana ko KilOnA xiyA _ROOT_ | k1 k4 k2 main

33. Parse - I xiyA KilOnAmohanarAma k2 k4 k1 _ROOT_ main

34. Integer Programming Constraints X ijk represents a possible arc from word group i to j with karaka label k It takes a value 1 if the solution has that arc and 0 otherwise. It cannot take any other values. The constraint rules are formulated into constraint equations.

35. Constraint Equations C1: For each demand group i, for each of its mandatory demands k, the following equalities must hold: M ik :  j x ikj = 1 C2: For each demand group i, for each of its optional or desirable demands k, the following inequalities must hold: O ik  :  j x ikj < = 1 C3: For each of the source groups j, the following equalities must hold: S j :  ik x ikj = 1

36. Multiple Frames If more than one karaka frame for a verb Call Integer Programming package for each frame If more than one demand groups (e.g., multiple verbs) in the sentence with multiple demand frames Call Integer Programming package for each combination of such frames

37. Other frames Common karaka frame Attached to each karaka frame Preference given to main frame if there are clashes Fallback karaka frame required karaka frame is missing Graceful degradation

38. Stage I: Types being handled Simple Verbs Non-finite verbs wA_huA wA_hI nA kara 0_rahe, etc. Copula Genitive

39. Example (Complex Sentence) rAma ne phala khaakara mohana ko Ram ‘ERG’ fruit ‘having eaten’ Mohan ‘DAT’ KilOnA xiyA toy gave ‘Having eaten the fruit Ram gave the toy to Mohan’

40. Candidates rAma ne phala khaakara mohana ko KilOnA xiyA _ROOT_ | X1: k1 X3: k2 X5: k4 X2: k2 X7: vmod X4: k2 X6: k2 X8: main

41. Constraint Equations Verb ‘xe’ Mandatory Demands (C1) k1  x1 = 1 k2  x2 + x3 + x4 = 1 Optional Demands (C2) k4  x5 <= 1 Verb ‘khaa’ Mandatory Demands (C1) k2  x6 = 1 vmod  x7 = 1 _ROOT_ C1 Main  x8 = 1

42. Constraint Equations (contd.) Incoming Arcs into Source (C3) rAma x1 = 1 phala x4 + x6 = 1 khaa x7 = 1 mohana x3 + x5 = 1 KilOnA x2 = 1 xe x8 = 1

43. Solution Graph xiyA KilOnA mohanarAma k2 k4 k1 _ROOT_ main khaakara phala k2 vmod

44. References Akshar Bharati and Rajeev Sangal. 1993. Parsing free word order languages in Paninian Framework. ACL:93, Proc.of Annual Meeting of Association of Computational Linguistics, Association of Computational Linguistics, New Jersey. USA. Akshar Bharati, Rajeev Sangal, T Papi Reddy. 2002. A Constraint Based Parser Using Integer Programming In Proc. of ICON-2002: International Conference on Natural Language Processing. Rafiya Begum, Samar Husain, Arun Dhwaj, Dipti Misra Sharma, Lakshmi Bai and Rajeev Sangal. 2008. Dependency Annotation Scheme for Indian Languages. In Proceedings of The Third International Joint Conference on Natural Language Processing (IJCNLP). Hyderabad, India. S. M. Shieber. 1985. Evidence against the context-freeness of natural language. In Linguistics and Philosophy, p. 8, 334–343. Tara Mohanan, 1994. Arguments in Hindi. CSLI Publications.

45. THANKS!!