NeuralLog: Natural Language Inference with Joint Neural and Logical Reasoning

Chen, Zeming; Gao, Qiyue; Moss, Lawrence S.

doi:10.18653/v1/2021.starsem-1.7

Cited by 18 publications

(13 citation statements)

References 29 publications

(44 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This parallels the increase in performance we observe from mid-training on semi-synthetic deduction steps, although they do not tackle the full pipeline task. Hu et al (2020) and Chen et al (2021b) propose systems which perform single-sentence natural language inference through proof search in the natural logic space. Explicitly using the monotonicity calculus (Icard et al, 2017) and natural logic to generate contrastive examples for semisynthetic mid-training is a promising future direction that could help to address the monotonicity issues encountered by our step deduction model, as discussed in Section 5.4.…”

Section: Related Workmentioning

confidence: 99%

Natural Language Deduction through Search over Statement Compositions

Bostrom¹,

Sprague²,

Chaudhuri³

et al. 2022

Preprint

View full text Add to dashboard Cite

In settings from fact-checking to question answering, we frequently want to know whether a collection of evidence entails a hypothesis. Existing methods primarily focus on end-toend discriminative versions of this task, but less work has treated the generative version in which a model searches over the space of entailed statements to derive the hypothesis. We propose a system for natural language deduction that decomposes the task into separate steps coordinated by best-first search, producing a tree of intermediate conclusions that faithfully reflects the system's reasoning process. Our experiments demonstrate that the proposed system can better distinguish verifiable hypotheses from unverifiable ones and produce natural language explanations that are more internally consistent than those produced by an end-to-end T5 model.

show abstract

Section: Related Workmentioning

confidence: 99%

Natural Language Deduction through Search over Statement Compositions

Bostrom¹,

Sprague²,

Chaudhuri³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…To discover potential tasks that can provide essential linguistic information for symbolic inferences, we studied four major logical systems for NLI, all with high accuracy on SICK (Marelli et al 2014), and several challenge datasets for the NLI task. They include NLI systems based on natural logic (Abzianidze 2020), monotonicity reasoning (Hu et al 2020;Chen, Gao, and Moss 2021), and theorem proving (Yanaka et al 2018).…”

Section: Inference Information Probesmentioning

confidence: 99%

“…This task probes the graph-based abstract meaning representation for sentences, a type of knowledge found effective in symbolic systems for acquiring paraphrase pairs and selecting correct inference steps (Yanaka et al 2018;Chen, Gao, and Moss 2021). The task is to construct a semantic graph that captures connections between concepts, modifiers, and relations in a sentence.…”

Section: Semantic Graph Construction (Semgraph)mentioning

confidence: 99%

Probing Linguistic Information for Logical Inference in Pre-trained Language Models

Chen

Gao

2022

AAAI

Self Cite

View full text Add to dashboard Cite

Progress in pre-trained language models has led to a surge of impressive results on downstream tasks for natural language understanding. Recent work on probing pre-trained language models uncovered a wide range of linguistic properties encoded in their contextualized representations. However, it is unclear whether they encode semantic knowledge that is crucial to symbolic inference methods. We propose a methodology for probing knowledge for inference that logical systems require but often lack in pre-trained language model representations. Our probing datasets cover a list of key types of knowledge used by many symbolic inference systems. We find that (i) pre-trained language models do encode several types of knowledge for inference, but there are also some types of knowledge for inference that are not encoded, (ii) language models can effectively learn missing knowledge for inference through fine-tuning. Overall, our findings provide insights into which aspects of knowledge for inference language models and their pre-training procedures capture. Moreover, we have demonstrated language models' potential as semantic and background knowledge bases for supporting symbolic inference methods.

show abstract

“…Natural logic dates back to the formalisms of Sanchez (1991), but has been received more recent treatments and reformulations in (MacCartney and Manning, 2007;Hu and Moss, 2018). Symbolic and hybrid neural/symbolic implementations of the natural logic paradigm have been explored in (Chen et al, 2021;Kalouli et al, 2020;Abzianidze, 2017;Hu et al, 2020).…”

Section: Related Workmentioning

confidence: 99%

Decomposing Natural Logic Inferences in Neural NLI

Rozanova¹,

Ferreira²,

Valentino³

et al. 2021

Preprint

View full text Add to dashboard Cite

In the interest of interpreting neural NLI models and their reasoning strategies, we carry out a systematic probing study which investigates whether these models capture the crucial semantic features central to natural logic: monotonicity and concept inclusion. Correctly identifying valid inferences in downwardmonotone contexts is a known stumbling block for NLI performance, subsuming linguistic phenomena such as negation scope and generalized quantifiers. To understand this difficulty, we emphasize monotonicity as a property of a context and examine the extent to which models capture monotonicity information in the contextual embeddings which are intermediate to their decision making process. Drawing on the recent advancement of the probing paradigm, we compare the presence of monotonicity features across various models. We find that monotonicity information is notably weak in the representations of popular NLI models which achieve high scores on benchmarks, and observe that previous improvements to these models based on fine-tuning strategies have introduced stronger monotonicity features together with their improved performance on challenge sets.

show abstract

NeuralLog: Natural Language Inference with Joint Neural and Logical Reasoning

Cited by 18 publications

References 29 publications

Natural Language Deduction through Search over Statement Compositions

Natural Language Deduction through Search over Statement Compositions

Probing Linguistic Information for Logical Inference in Pre-trained Language Models

Decomposing Natural Logic Inferences in Neural NLI

Contact Info

Product

Resources

About