While recent advancements in large language models (LLMs) bring us closer to achieving artificial general intelligence, the question persists: Do LLMs truly understand language, or do they merely mimic comprehension through pattern recognition? This study seeks to explore this question through the lens of syntax, a crucial component of sentence comprehension. Adopting a natural language question-answering (Q&A) scheme, we craft questions targeting nine syntactic knowledge points that are most closely related to sentence comprehension. Experiments conducted on 24 LLMs suggest that most have a limited grasp of syntactic knowledge, exhibiting notable discrepancies across different syntactic knowledge points. In particular, questions involving prepositional phrase attachment pose the greatest challenge, whereas those concerning adjectival modifier and indirect object are relatively easier for LLMs to handle. Furthermore, a case study on the training dynamics of the LLMs reveals that the majority of syntactic knowledge is learned during the initial stages of training, hinting that simply increasing the number of training tokens may not be the `silver bullet' for improving the comprehension ability of LLMs.
We revisit the challenging problem of resolving prepositional-phrase (PP) attachment ambiguity. To date, proposed solutions are either rule-based, where explicit grammar rules direct how to resolve ambiguities; or statistical, where the decision is learned from a corpus of labeled examples. We argue that explicit commonsense knowledge bases can provide an essential ingredient for making good attachment decisions. We implemented a module, named Patch-Comm, that can be used by a variety of conventional parsers, to make attachment decisions. Where the commonsense KB does not provide direct answers, we fall back on a more general system that infers "out-of-knowledge-base" assertions in a manner similar to the way some NLP systems handle out-of-vocabulary words. Our results suggest that the commonsense knowledge-based approach can provide the best of both worlds, integrating rule-based and statistical techniques. As the field is increasingly coming to recognize the importance of explainability in AI, a commonsense approach can enable NLP developers to better understand the behavior of systems, and facilitate natural dialogues with end users.
We show that in language learning, contrary to received wisdom, keeping exceptional training instances in memory can be beneficial for generalization accuracy. We investigate this phenomenon empirically on a selection of benchmark natural language processing tasks: grapheme-to-phoneme conversion, part-of-speech tagging, prepositional-phrase attachment, and base noun phrase chunking. In a first series of experiments we combine memory-based learning with training set editing techniques, in which instances are edited based on their typicality and class prediction strength. Results show that editing exceptional instances (with low typicality or low class prediction strength) tends to harm generalization accuracy. In a second series of experiments we compare memory-based learning and decision-tree learning methods on the same selection of tasks, and find that decision-tree learning often performs worse than memory-based learning. Moreover, the decrease in performance can be linked to the degree of abstraction from exceptions (i.e., pruning or eagerness). We provide explanations for both results in terms of the properties of the natural language processing tasks and the learning algorithms.
An important characteristic of English written text is the abundance of noun compounds - sequences of nouns acting as a single noun, e.g., colon cancer tumor suppressor protein. While eventually mastered by domain experts, their interpretation poses a major challenge for automated analysis. Understanding noun compounds' syntax and semantics is important for many natural language applications, including question answering, machine translation, information retrieval, and information extraction. I address the problem of noun compounds syntax by means of novel, highly accurate unsupervised and lightly supervised algorithms using the Web as a corpus and search engines as interfaces to that corpus. Traditionally the Web has been viewed as a source of page hit counts, used as an estimate for n-gram word frequencies. I extend this approach by introducing novel surface features and paraphrases, which yield state-of-the-art results for the task of noun compound bracketing. I also show how these kinds of features can be applied to other structural ambiguity problems, like prepositional phrase attachment and noun phrase coordination. I address noun compound semantics by automatically generating paraphrasing verbs and prepositions that make explicit the hidden semantic relations between the nouns in a noun compound. I also demonstrate how these paraphrasing verbs can be used to solve various relational similarity problems, and how paraphrasing noun compounds can improve machine translation.
The common practice in coreference resolution is to identify and evaluate the maximum span of mentions. The use of maximum spans tangles coreference evaluation with the challenges of mention boundary detection like prepositional phrase attachment. To address this problem, minimum spans are manually annotated in smaller corpora. However, this additional annotation is costly and therefore, this solution does not scale to large corpora. In this paper, we propose the MINA algorithm for automatically extracting minimum spans to benefit from minimum span evaluation in all corpora. We show that the extracted minimum spans by MINA are consistent with those that are manually annotated by experts. Our experiments show that using minimum spans is in particular important in cross-dataset coreference evaluation, in which detected mention boundaries are noisier due to domain shift. We will integrate MINA into https://github.com/ns-moosavi/coval for reporting standard coreference scores based on both maximum and automatically detected minimum spans.
We present an approach to simultaneously perform semantic segmentation and prepositional phrase attachment resolution for captioned images. Some ambiguities in language cannot be resolved without simultaneously reasoning about an associated image. If we consider the sentence "I shot an elephant in my pajamas", looking at language alone (and not using common sense), it is unclear if it is the person or the elephant wearing the pajamas or both. Our approach produces a diverse set of plausible hypotheses for both semantic segmentation and prepositional phrase attachment resolution that are then jointly reranked to select the most consistent pair. We show that our semantic segmentation and prepositional phrase attachment resolution modules have complementary strengths, and that joint reasoning produces more accurate results than any module operating in isolation. Multiple hypotheses are also shown to be crucial to improved multiple-module reasoning. Our vision and language approach significantly outperforms the Stanford Parser (De Marneffe et al., 2006) by 17.91% (28.69% relative) and 12.83% (25.28% relative) in two different experiments. We also make small improvements over DeepLab-CRF (Chen et al., 2015).
Intelligent systems capable of automatically understanding natural language text are important for many artificial intelligence applications including mobile phone voice assistants, computer vision, and robotics. Understanding language often constitutes fitting new information into a previously acquired view of the world. However, many machine reading systems rely on the text alone to infer its meaning. In this paper, we pursue a different approach; machine reading methods that make use of background knowledge to facilitate language understanding. To this end, we have developed two methods: The first method addresses prepositional phrase attachment ambiguity. It uses background knowledge within a semi-supervised machine learning algorithm that learns from both labeled and unlabeled data. This approach yields state-of-the-art results on two datasets against strong baselines; The second method extracts relationships from compound nouns. Our knowledge-aware method for compound noun analysis accurately extracts relationships and significantly outperforms a baseline that does not make use of background knowledge.
Type-level word embeddings use the same set of parameters to represent all instances of a word regardless of its context, ignoring the inherent lexical ambiguity in language. Instead, we embed semantic concepts (or synsets) as defined in WordNet and represent a word token in a particular context by estimating a distribution over relevant semantic concepts. We use the new, context-sensitive embeddings in a model for predicting prepositional phrase(PP) attachments and jointly learn the concept embeddings and model parameters. We show that using context-sensitive embeddings improves the accuracy of the PP attachment model by 5.4% absolute points, which amounts to a 34.4% relative reduction in errors.
Neural machine translation (NMT) models are able to partially learn syntactic information from sequential lexical information. Still, some complex syntactic phenomena such as prepositional phrase attachment are poorly modeled. This work aims to answer two questions: 1) Does explicitly modeling target language syntax help NMT? 2) Is tight integration of words and syntax better than multitask training? We introduce syntactic information in the form of CCG supertags in the decoder, by interleaving the target supertags with the word sequence. Our results on WMT data show that explicitly modeling target-syntax improves machine translation quality for German->English, a high-resource pair, and for Romanian->English, a low-resource pair and also several syntactic phenomena including prepositional phrase attachment. Furthermore, a tight coupling of words and syntax improves translation quality more than multitask training. By combining target-syntax with adding source-side dependency labels in the embedding layer, we obtain a total improvement of 0.9 BLEU for German->English and 1.2 BLEU for Romanian->English.