AbstractWe suggest that foundation models are general purpose solutions similar to general purpose programmable microprocessors, where fine-tuning and prompt-engineering are analogous to coding for microprocessors. Evaluating general purpose solutions is not like hypothesis testing. We want to know how well the machine will perform on an unknown program with unknown inputs for unknown users with unknown budgets and unknown utility functions. This paper is based on an invited talk by John Mashey, “Lessons from SPEC,” at an ACL-2021 workshop on benchmarking. Mashey started by describing Standard Performance Evaluation Corporation (SPEC), a benchmark that has had more impact than benchmarks in our field because SPEC addresses an import commercial question: which CPU should I buy? In addition, SPEC can be interpreted to show that CPUs are 50,000 faster than they were 40 years ago. It is remarkable that we can make such statements without specifying the program, users, task, dataset, etc. It would be desirable to make quantitative statements about improvements of general purpose foundation models over years/decades without specifying tasks, datasets, use cases, etc.
AbstractIt’s less than a year since OpenAI’s board voted to fire Sam Altman as CEO, in a palace coup that lasted just a weekend before Altman was reinstated. That weekend and subsequent events in OpenAI’s storyline provide all the ingredients for a soap opera. So, just in case Netflix is interested, here’s a stab at a synopsis of what might be just the first of many seasons of ‘The Generative AI Wars’.
AbstractRetrieval-augmented generation (RAG) adds a simple but powerful feature to chatbots, the ability to upload files just-in-time. Chatbots are trained on large quantities of public data. The ability to upload files just-in-time makes it possible to reduce hallucinations by filling in gaps in the knowledge base that go beyond the public training data such as private data and recent events. For example, in a customer service scenario, with RAG, we can upload your private bill and then the bot can discuss questions about your bill as opposed to generic FAQ questions about bills in general. This tutorial will show how to upload files and generate responses to prompts; see https://github.com/kwchurch/RAG for multiple solutions based on tools from OpenAI, LangChain, HuggingFace transformers and VecML.
Randomized prospective studies represent the gold standard for experimental design. In this paper, we present a randomized prospective study to validate the benefits of combining rule-based and data-driven natural language understanding methods in a virtual patient dialogue system. The system uses a rule-based pattern matching approach together with a machine learning (ML) approach in the form of a text-based convolutional neural network, combining the two methods with a simple logistic regression model to choose between their predictions for each dialogue turn. In an earlier, retrospective study, the hybrid system yielded a nearly 50% error reduction on our initial data, in part due to the differential performance between the two methods as a function of label frequency. Given these gains, and considering that our hybrid approach is unique among virtual patient systems, we compare the hybrid system to the rule-based system by itself in a randomized prospective study. We evaluate 110 unique medical student subjects interacting with the system over 5,296 conversation turns, to verify whether similar gains are observed in a deployed system. This prospective study broadly confirms the findings from the earlier one but also highlights important deficits in our training data. The hybrid approach still improves over either rule-based or ML approaches individually, even handling unseen classes with some success. However, we observe that live subjects ask more out-of-scope questions than expected. To better handle such questions, we investigate several modifications to the system combination component. These show significant overall accuracy improvements and modest F1 improvements on out-of-scope queries in an offline evaluation. We provide further analysis to characterize the difficulty of the out-of-scope problem that we have identified, as well as to suggest future improvements over the baseline we establish here.
This paper proposes a sequence-to-sequence model for data-to-text generation, called DM-NLG, to generate a natural language text from structured nonlinguistic input. Specifically, by adding a dynamic memory module to the attention-based sequence-to-sequence model, it can store the information that leads to generate previous output words and use it to generate the next word. In this way, the decoder part of the model is aware of all previous decisions, and as a result, the generation of duplicate words or incomplete semantic concepts is prevented. To improve the generated sentences quality by the DM-NLG decoder, a postprocessing step is performed using the pretrained language models. To prove the effectiveness of the DM-NLG model, we performed experiments on five different datasets and observed that our proposed model is able to reduce the slot error rate rate by 50% and improve the BLEU by 10%, compared to the state-of-the-art models.
Recently, neural abstractive text summarization (NATS) models based on sequence-to-sequence architecture have drawn a lot of attention. Real-world texts that need to be summarized range from short news with dozens of words to long reports with thousands of words. However, most existing NATS models are not good at summarizing long documents, due to the inherent limitations of their underlying neural architectures. In this paper, we focus on the task of long document summarization (LDS). Based on the inherent section structures of source documents, we divide an abstractive LDS problem into several smaller-sized problems. In this circumstance, how to provide a less-biased target summary as the supervision for each section is vital for the model’s performance. As a preliminary, we formally describe the section-to-summary-sentence (S2SS) alignment for LDS. Based on this, we propose a novel NATS framework for the LDS task. Our framework is built based on the theory of unbalanced optimal transport (UOT), and it is named as UOTSumm. It jointly learns three targets in a unified training objective, including the optimal S2SS alignment, a section-level NATS summarizer, and the number of aligned summary sentences for each section. In this way, UOTSumm directly learns the text alignment from summarization data, without resorting to any biased tool such as ROUGE. UOTSumm can be easily adapted to most existing NATS models. And we implement two versions of UOTSumm, with and without the pretrain-finetune technique. We evaluate UOTSumm on three publicly available LDS benchmarks: PubMed , arXiv , and GovReport . UOTSumm obviously outperforms its counterparts that use ROUGE for the text alignment. When combined with UOTSumm, the performance of two vanilla NATS models improves by a large margin. Besides, UOTSumm achieves better or comparable performance when compared with some recent strong baselines.
AbstractAbstractive summarization is an approach to document summarization that is not limited to selecting sentences from the document but can generate new sentences as well. We address the two main challenges in abstractive summarization: how to evaluate the performance of a summarization model and what is a good training objective. We first introduce new evaluation measures based on the semantic similarity of the input and corresponding summary. The similarity scores are obtained by the fine-tuned BERTurk model using either the cross-encoder or a bi-encoder architecture. The fine-tuning is done on the Turkish Natural Language Inference and Semantic Textual Similarity benchmark datasets. We show that these measures have better correlations with human evaluations compared to Recall-Oriented Understudy for Gisting Evaluation (ROUGE) scores and BERTScore. We then introduce a deep reinforcement learning algorithm that uses the proposed semantic similarity measures as rewards, together with a mixed training objective, in order to generate more natural summaries in terms of human readability. We show that training with a mixed training objective function compared to only the maximum-likelihood objective improves similarity scores.
Web search is an experience that naturally lends itself to recommendations, including query suggestions and related entities. In this article, we propose to recommend specific tasks to users, based on their search queries, such as planning a holiday trip or organizing a party. Specifically, we introduce the problem of query-based task recommendation and develop methods that combine well-established term-based ranking techniques with continuous semantic representations, including sentence representations from several transformer-based models. Using a purpose-built test collection, we find that our method is able to significantly outperform a strong text-based baseline. Further, we extend our approach to using a set of queries that all share the same underlying task, referred to as search mission, as input. The study is rounded off with a detailed feature and query analysis.
Table-to-text generation aims to generate descriptions for structured data (i.e., tables) and has been applied in many fields like question-answering systems and search engines. Current approaches mostly use neural language models to learn alignment between output and input based on the attention mechanisms, which are still flawed by the gradual weakening of attention when processing long texts and the inability to utilize the records' structural information. To solve these problems, we propose a novel generative model SAN-T2T, which consists of a field-content selective encoder and a descriptive decoder, connected with a selective attention network. In the encoding phase, the table's structure is integrated into its field representation, and a content selector with self-aligned gates is applied to take advantage of the fact that different records can determine each other's importance. In the decoding phase, the content selector's semantic information enhances the alignment between description and records, and a featured copy mechanism is applied to solve the rare word problem. Experiments on WikiBio and WeatherGov datasets show that SAN-T2T outperforms the baselines by a large margin, and the content selector indeed improves the model's performance.
The technical and mainstream media's headline coverage of AI invariably centers around the often astounding abilities demonstrated by large language models. That's hardly surprising, since to all intents and purposes that's where the newsworthy magic of generative AI lies. But it takes a village to raise a child: behind the scenes, there's an entire ecosystem that supports the development and deployment of these models and the applications that are built on top of them. Some parts of that ecosystem are dominated by the Big Tech incumbents, but there are also many niches where start-ups are aiming to gain a foothold. We take a look at some components of that ecosystem, with a particular focus on ideas that have led to investment in start-ups over the last year or so.
This article discusses the development of an automated plot extraction system for narrative texts. Acknowledging the distinction between plot, as an object of study with its own rich history and literature, and features of a text that may be automatically extractable, we begin by characterizing a text's scatter plot of entities. This visualization of a text reveals entity density patterns characterizing the particular telling of the story under investigation and leads to effective scene partitioning. We then introduce the concept of narrative flow, a graph representation of the narrative ordering of scenes (the syuzhet) that includes how entities move through scenes from the text, and investigate the degree to which narrative flow can be automatically extracted given a glossary of plot-important objects, actors, and locations. Our subsequent analysis then explores the correlation between subjective notions of plot and the information extracted through these visualizations. In particular, we discuss narrative structures commonly found within the graphs and make comparisons with ground truth narrative flow graphs, showing mixed results highlighting the difficulty of plot extraction. However, the visual artifacts and common structural relationships seen in the graphs provide insight into narrative and its underlying plot.
In this paper, we propose a novel way of improving named entity recognition (NER) in the Korean language using its language-specific features. While the field of NER has been studied extensively in recent years, the mechanism of efficiently recognizing named entities (NEs) in Korean has hardly been explored. This is because the Korean language has distinct linguistic properties that present challenges for modeling. Therefore, an annotation scheme for Korean corpora by adopting the CoNLL-U format, which decomposes Korean words into morphemes and reduces the ambiguity of NEs in the original segmentation that may contain functional morphemes such as postpositions and particles, is proposed herein. We investigate how the NE tags are best represented in this morpheme-based scheme and implement an algorithm to convert word-based and syllable-based Korean corpora with NEs into the proposed morpheme-based format. Analyses of the results of traditional and neural models reveal that the proposed morpheme-based format is feasible, and the varied performances of the models under the influence of various additional language-specific features are demonstrated. Extrinsic conditions were also considered to observe the variance of the performances of the proposed models, given different types of data, including the original segmentation and different types of tagging formats.
Usage of large language models and chat bots will almost surely continue to grow, since they are so easy to use, and so (incredibly) credible. I would be more comfortable with this reality if we encouraged more evaluations with humans-in-the-loop to come up with a better characterization of when the machine can be trusted and when humans should intervene. This article will describe a homework assignment, where I asked my students to use tools such as chat bots and web search to write a number of essays. Even after considerable discussion in class on hallucinations, many of the essays were full of misinformation that should have been fact-checked. Apparently, it is easier to believe ChatGPT than to be skeptical. Fact-checking and web search are too much trouble.
We examine the binary classification of sentiment views for verbal multiword expressions (MWEs). Sentiment views denote the perspective of the holder of some opinion. We distinguish between MWEs conveying the view of the speaker of the utterance (e.g., in "The company reinvented the wheel" the holder is the implicit speaker who criticizes the company for creating something already existing) and MWEs conveying the view of explicit entities participating in an opinion event (e.g., in "Peter threw in the towel" the holder is Peter having given up something). The task has so far been examined on unigram opinion words. Since many features found effective for unigrams are not usable for MWEs, we propose novel ones taking into account the internal structure of MWEs, a unigram sentiment-view lexicon and various information from Wiktionary. We also examine distributional methods and show that the corpus on which a representation is induced has a notable impact on the classification. We perform an extrinsic evaluation in the task of opinion holder extraction and show that the learnt knowledge also improves a state-of-the-art classifier trained on BERT. Sentiment-view classification is typically framed as a task in which only little labeled training data are available. As in the case of unigrams, we show that for MWEs a feature-based approach beats state-of-the-art generic methods.
Despite recent breakthroughs in Machine Learning for Natural Language Processing, the Natural Language Inference (NLI) problems still constitute a challenge. To this purpose we contribute a new dataset that focuses exclusively on the factivity phenomenon; however, our task remains the same as other NLI tasks, i.e. prediction of entailment, contradiction or neutral (ECN). The dataset contains entirely natural language utterances in Polish and gathers 2,432 verb-complement pairs and 309 unique verbs. The dataset is based on the National Corpus of Polish (NKJP) and is a representative sample in regards to frequency of main verbs and other linguistic features (e.g. occurrence of internal negation). We found that transformer BERT-based models working on sentences obtained relatively good results ($\approx89\%$ F1 score). Even though better results were achieved using linguistic features ($\approx91\%$ F1 score), this model requires more human labour (humans in the loop) because features were prepared manually by expert linguists. BERT-based models consuming only the input sentences show that they capture most of the complexity of NLI/factivity. Complex cases in the phenomenon - e.g. cases with entitlement (E) and non-factive verbs - remain an open issue for further research.
We propose a multilingual data-driven method for generating reading comprehension questions using dependency trees. Our method provides a strong, mostly deterministic, and inexpensive-totrain baseline for less-resourced languages. While a language-specific corpus is still required, its size is nowhere near those required by modern neural question generation (QG) architectures. Our method surpasses QG baselines previously reported in the literature and shows a good performance in terms of human evaluation.
In this paper, we study natural language paraphrasing from both corpus creation and modeling points of view. We focus in particular on the methodology that allows the extraction of challenging examples of paraphrase pairs in their natural textual context, leading to a dataset potentially more suitable for evaluating the models’ ability to represent meaning, especially in document context, when compared with those gathered using various sentence-level heuristics. To this end, we introduce the Turku Paraphrase Corpus, the first large-scale, fully manually annotated corpus of paraphrases in Finnish. The corpus contains 104,645 manually labeled paraphrase pairs, of which 98% are verified to be true paraphrases, either universally or within their present context. In order to control the diversity of the paraphrase pairs and avoid certain biases easily introduced in automatic candidate extraction, the paraphrases are manually collected from different paraphrase-rich text sources. This allows us to create a challenging dataset including longer and more lexically diverse paraphrases than can be expected from those collected through heuristics. In addition to quality, this also allows us to keep the original document context for each pair, making it possible to study paraphrasing in context. To our knowledge, this is the first paraphrase corpus which provides the original document context for the annotated pairs. We also study several paraphrase models trained and evaluated on the new data. Our initial paraphrase classification experiments indicate a challenging nature of the dataset when classifying using the detailed labeling scheme used in the corpus annotation, the accuracy substantially lacking behind human performance. However, when evaluating the models on a large scale paraphrase retrieval task on almost 400M candidate sentences, the results are highly encouraging, 29–53% of the pairs being ranked in the top 10 depending on the paraphrase type. The Turku Paraphrase Corpus is available at github.com/TurkuNLP/Turku-paraphrase-corpus as well as through the popular HuggingFace datasets under the CC-BY-SA license.
AbstractWord order is one of the most important grammatical devices and the basis for language understanding. However, as one of the most popular NLP architectures, Transformer does not explicitly encode word order. A solution to this problem is to incorporate position information by means of position encoding/embedding (PE). Although a variety of methods of incorporating position information have been proposed, the NLP community is still in want of detailed statistical researches on position information in real-life language. In order to understand the influence of position information on the correlation between words in more detail, we investigated the factors that affect the frequency of words and word sequences in large corpora. Our results show that absolute position, relative position, being at one of the two ends of a sentence and sentence length all significantly affect the frequency of words and word sequences. Besides, we observed that the frequency distribution of word sequences over relative position carries valuable grammatical information. Our study suggests that in order to accurately capture word–word correlations, it is not enough to focus merely on absolute and relative position. Transformers should have access to more types of position-related information which may require improvements to the current architecture.
Automatic paraphrase detection is the task of measuring the semantic overlap between two given texts. A major hurdle in the development and evaluation of paraphrase detection approaches, particularly for South Asian languages like Urdu, is the inadequacy of standard evaluation resources. The very few available paraphrased corpora for these languages are manually created. As a result, they are constrained to smaller sizes and are not very feasible to evaluate mainstream data-driven and deep neural networks (DNNs)-based approaches. Consequently, there is a need to develop semi-or fully automated corpus generation approaches for the resource-scarce languages. There is currently no semi-or fully automatically generated sentence-level Urdu paraphrase corpus. Moreover, no study is available to localize and compare approaches for Urdu paraphrase detection that focus on various mainstream deep neural architectures and pretrained language models.This research study addresses this problem by presenting a semi-automatic pipeline for generating paraphrased corpora for Urdu. It also presents a corpus that is generated using the proposed approach. This corpus contains 3147 semi-automatically extracted Urdu sentence pairs that are manually tagged as paraphrased (854) and non-paraphrased (2293). Finally, this paper proposes two novel approaches based on DNNs for the task of paraphrase detection in Urdu text. These are Word Embeddings n-gram Overlap (henceforth called WENGO), and a modified approach, Deep Text Reuse and Paraphrase Plagiarism Detection (henceforth called D-TRAPPD). Both of these approaches have been evaluated on two related tasks: (i) paraphrase detection, and (ii) text reuse and plagiarism detection. The results from these evaluations revealed that D-TRAPPD (F-1 = 96.80 for paraphrase detection and F-1 = 88.90 for text reuse and plagiarism detection) outperformed WENGO (F-1 = 81.64 for paraphrase detection and F-1 = 61.19 for text reuse and plagiarism detection) as well as other state-of-the-art approaches for these two tasks. The corpus, models, and our implementations have been made available as free to download for the research community.
