Ableitung von Information aus Texten – Textual Entailment
Kursbeschreibung
| Studiengang | Modulkürzel | Leistungs- bewertung |
|---|---|---|
| BA-2010 | AS-CL, AS-FL | 8 LP |
| BA-2010[100%|75%] | CS-CL | 6 LP |
| BA-2010[50%] | BS-CL | 6 LP |
| BA-2010[25%] | BS-AC, BS-FL | 4 LP |
| Master | SS-CL-TAC, SS-SC-FAL | 8 LP |
| Dozent | Kurt Eberle |
| Veranstaltungsart | |
| Sprache | Deutsch/Englisch |
| Erster Termin | 10.02.2025 |
| Letzter Termin | 14.02.2025 |
| Zeit und Ort | 09:00 - 14:30, INF 325 / SR24 |
| Commitment-Frist | tba. |
Achtung: Anmeldung zur Seminar-Teilnahme bis 20.01.2025!
per email oder in: Moodle
Teilnehmerkreis/Participants
All advanced Bachelor students and all Master students.
Students from Computer Science, Mathematics or Scientific computing
with Anwendungsgebiet Computational Linguistics are welcome.
Teilnahmevoraussetzungen/Prerequisites for Participation
Teilnehmerkreis/Participants
All advanced Bachelor students and all Master students. Students from Computer Science, Mathematics or Scientific computing with Anwendungsgebiet Computational Linguistics are welcome.Teilnahmevoraussetzungen/Prerequisites for Participation
Introduction to Computational Linguistics or similar introductory courses
Basic knowledge in:
- Logic (Logische Grundlagen für die Computerlinguistik) and
- Formal Semantics (Logic representations of texts, Discourse Representation Theory or similar)
Leistungsnachweis/Assessment
- Paper presentation (4/2 LP)
- Written Exam (4 LP)
Inhalt/Content
When does a text follow from another? In traditional formal semantics, this is checked with the help of semantic representations: The representations of the two texts are computed, typically within the framework of a modeling system based on Montague grammar, and then it is checked whether the representation of the second text can be inferred from that of the first by using the deduction rules of the system.
It is not a new finding that statements that people typically derive from texts often go beyond what logical inference can deliver. Formal semantic attempts have been made to model such phenomena via 'abduction' and various 'default logics', etc.
Since more and more phenomena in computational linguistics have been successfully modeled with machine learning (ML) approaches, such methods have been tried for semantic inference also. In this sense, 'textual entailment' means 'learning human inference behavior from data'; from data that consist of pairs of (short) texts and even shorter statements (hypotheses) and judgements about whether the hypotheses follow from the texts respectively.
In the seminar we will take a look at various methods that have been suggested for deriving semantic information from texts and for relating corresponding pieces of information to each other; where we will contrast traditional rule-based methods with modern data-driven methods.
We will start with a brief review of different types of deep and shallow text representations and an overview of corresponding representation-specific inference phenomena. Then we consider the generation of semantic representations, where we contrast formal versus ML means. The main part of the seminar will then be devoted to a range of entailment approaches with and without representations, with an emphasis on recent ML-based approaches, including recent methods with large language models such as BERT or in ChatGPT
The goal is to gain a good overview of recent work on the topic, based on a good formal understanding of the phenomena.
Vorraussichtlicher Kursplan/Agenda
| Montag | Slides | ||
| 9.15 | Introduction | Organisation, Motivation: Text entailment, text and hypothesis, representations(do we need reps?), Program, Pascal Development Testsuite (PDT) (Dagan,Glickman, Magnini 2006) | Intro |
| 11.00 | Types of Inferences | Entailment , Conventional & Conversational Implicature and the Pascal Development Suite (Zaenen, Karttunen, Crouch 2005): What is in PDT und what should be there?' | Intro |
| 13.00 | Sem. Representations | TE with deep vs shallow representation: Predicate logic vs 'Light-weight semantics' (Blackburn Bos 2003) vs (Monz de Rijke 2001) | Intro |
| Tuesday | |||
| Semantic construction, corpora, TE systems | |||
| Wednesday | |||
| TE with rules vs TE with features and statistics, neural nets | |||
| Thursday | |||
| Approaches with NNs, different models | |||
| Friday | |||
| Others?, difficulties, discourse relations, summary | |||
Literatur/Literature
(Overview. Of course, only a rather small part of this will be discussed in the seminar)
| 1) | Phänomene und Repräsentation | |
| Blackburn, Bos 2003 | Computational Semantics. Theoria 18(1): 27-45 | |
| Condoravdi, Crouch et al. 2003 | Entailment, intensionality and text understanding | |
| de Marneffe, Rafferty, Manning 2008 | Finding contradictions in text | |
| Sánchez Valencia 1991 | Studies on Natural Logic and Categorial Grammar. Ph.D. thesis, Univ. of Amsterdam | |
| Fyodorov, Winter, Francez 2000 | A natural logic inference system | |
| Lakoff 1970 | Linguistics and Natural Logic | |
| MacCartney Manning 2008a | Natural Logic for Textual Inference | |
| MacCartney Manning 2008b | Modeling semantic Containment and Exclusion in Natural Language Inference | |
| MacCartney Manning 2009 | An extended model of natural logic | |
| Mronz C., de Rijke 2001 | Light-Weight Entailment Checking for Computational Semantics | |
| Pinkal M., 2007 | Seminar on Entailment | |
| Zaenen, Karttunen, Crouch 2005 | Local Textual Inference: Can it be Defined or Circumscribed? in: ACL 2005 | |
| 2) | Semantic Parsing | |
| Conneau, A et al. 2017 | Supervised learning of universal sentence representations from natural language inference data | |
| Liang Potts 2014 | Bringing machine learning and compositional semantics together | |
| Liang 2016 | Learning executable semantic parsers for natural language understanding | |
| Zettlemoyer, Collins 2012 | Learning to Map Sentences to Logical Form: Structured Classification with Probabilistic Categorial Grammars | |
| 3) | Corpora, (Parallel) Meaning Banks | |
| https://aclweb.org/aclwiki/Textual_Entailment_Resource_Pool | ||
| Abzianidze et al. 2017 | The Parallel Meaning Bank: Towards a Multilingual Corpus of Translations Annotated with Compositional Meaning Representations | |
| Banarescu et al. 2013 | Abstract Meaning Representation for sembanking | |
| Bentivogli et al. 2010 | Building Textual Entailment Specialized Data Sets | |
| Bos et al. 2017 | The Groningen Meaning Bank | |
| Bowman, Angeli, Potts, Manning, 2015 | A large annotated corpus for learning natural language inference | |
| Cooper et al. 1996 | THE FRACAS TEXTUAL INFERENCE PROBLEM SET | |
| Williams et al. 2018 | A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference | |
| 4) | Textual Entailment | |
| Androutsopoulos, Malakasiotis 2010 | A Survey of Paraphrasing and Textual Entailment Methods | |
| Bentivogli et al. 2009 | The fifth Pascal recognizing Textual Entailment Challenge | |
| Bos , Markert. 2005 | Recognising textual entailment with logical inference in: HLT/EMNLP 2005 | |
| Bos, Markert 2006 | Combining Shallow and Deep NLP Methods for Recognizing Textual Entailment | |
| Bowman, Gauthier, Rastogi, Gupta, Manning, Potts 2016 | A fast unified model for parsing and sentence understanding | |
| Burchardt A | Modeling Textual Entailment with Role-Semantic Information | |
| Cabrio Magnini 2010 | Towards Qualitative Entailment of Textual Entailment Systems | |
| Cabrio, Magnini 2009 | Defining specialized entailment engines using natural logic relations. in Vetulani: Human Language Technology | |
| Chambers et al. 2007 | Learning Alignments and Leveraging Natural Logic | |
| Chen, Zhu, Ling, Wei, Jiang, Inkpen 2017 | Enhanced LSTM for natural language inference | |
| Dagan, Glickman 2004 | Probabilistic Textual Entailment: Generic Applied Modeling of Language Variability | |
| Dagan, Dolan et al. 2009 | Recognizing Textual Entailment: Rational, evaluation and approaches | |
| Dagan, Glickman, Magnini 2006 | The PASCAL recognising textual entailment challenge. In MLCW pp 177-190 | |
| de Salvo Braz et al. 2005 | An Inference Model for Semantic Entailment in Natural Language | |
| Kouylekov Magnini 2005 | Tree-Edit Distance for Textual Entailment | |
| Lien, Kouylekov 2015 | Semantic Parsing for Textual Entailment | |
| MacCartney 2009 | Natural language inference | |
| MacCartney et al. 2006 | Learning to recognize features of valid textual entailments | |
| Marelli et al. 2014 | SemEval-2014 Task 1: Evaluation of compositional distributional semantic models on full sentences through semantic relatedness and textual entailment | |
| Mirkin, Dagan, Padó 2010 | Assessing the Role of Discourse References in Entailment Inference | |
| Monz 1999 | Contextual Inference in Computational Semantics | |
| Monz, de Rijke 2001 | Light-Weight Entailment Checking for Computational Semantics | |
| Mou, Men, Li, Xu, Zhang, Yan,Jin. 2016 | Natural language inference by tree-based convolution and heuristic matching | |
| Nangia, Williams, Lazaridou, Bowman 2017 | The RepEval 2017 Shared Task: Multi-genre natural language inference with sentence representations. repeval01 | |
| Nairn et al. 2006 | Computing Relative Polarity for Textual Entailment | |
| Pazienza, Pennacchiotti, Zanzotto 2005 | Learning Textual Entailment on a Distance Feature Space in : MLWC pp 240-260 | |
| Parikh, Täckström, Das Uszkoreit 2016 | A decomposable attention model for natural language inference | |
| Pérez, Alfonseca 2005 | Using Bleu-like Algorithms for the Automatic Recognition of Entailment in: MLCW, p 191 et seq. | |
| Quiñonero-Candela, Dagan et al 2005 | Machine Learning Challenges:
Evaluating Predictive Uncertainty, Visual Object Classification and Recognizing Textual Entailment First PASCAL Machine Learning Challenges Workshop, MLCW 2005(MLCW) |
|
| Storks,Gao, Chai 2020 | Recent Advances in Natural Language Inference: A Survey of Benchmarks, Resources, and Approaches | |
| Szpektor et al. 2007 | Instance-based Evaluation of Entailment Rule Acquisition | |
| TE References 2017 | ACLWeb list of references | |
| Vanderwende, Dolan 2005 | What Syntax Can Contribute in the Entailment Task in: MLWC p 205 et seq. | |
| Wang , Jiang 2016 | Learning natural language inference with LSTM | |
| 4a) | Textual Entailment with LLMs | |
| Tuteja, H. 2019 | Textual Entailment using Bert (Implementierung auf github) | Link |
| Alsuhaibani, M. 2023 | Deep Learning-based Sentence Embeddings using BERT for Textual Entailment | Link |
| Verma, Dh. 2021 | Fine-tuning pre-trained transformer models for sentence entailment. A PyTorch and Hugging Face implementation of fine-tuning BERT on the MultiNLI dataset | Link |
| ... | ... | ... |
