Thailand

This paper presents our submission to the L+M-24 shared task, focused on translating molecular structures into natural language descriptions, known as the molecule captioning task. We selected a small language model (SLM), Phi-3-mini-4k, to evaluate the impact of continued pretraining and instruction tuning for domain-specific chemical knowledge. The Phi-3 model was continued pretrained with 90M chemistry textbooks and abstracts, followed by instruction tuning on 150K question answering sets of SMILES and general chemistry knowledge. Despite the continued pretraining phase not including direct exposure to SMILES representations, it significantly enhanced the Phi-3 model&#39;s performance, a 300% increase for the BLEU scores, in the molecule captioning task. The code and model are released at \url{https://github.com/bluesky333/Phi3KnowChem} to facilitate research in chemical small language modeling.

ACL 2024

Knowlab&#39;s Submission to L+M Shared Task: All you need is continued pretraining of \chemistry texts even for molecule captioning

small language model

molecule

large language model

This paper presents our submission to the L+M-24 shared task, focused on translating molecular structures into natural language descriptions, known as the molecule captioning task. We selected a small language model (SLM), Phi-3-mini-4k, to evaluate the impact of continued pretraining and instruction tuning for domain-specific chemical knowledge. The Phi-3 model was continued pretrained with 90M chemistry textbooks and abstracts, followed by instruction tuning on 150K question answering sets of SMILES and general chemistry knowledge. Despite the continued pretraining phase not including direct exposure to SMILES representations, it significantly enhanced the Phi-3 model's performance, a 300% increase for the BLEU scores, in the molecule captioning task. The code and model are released at \url{https://github.com/bluesky333/Phi3KnowChem} to facilitate research in chemical small language modeling.

Knowlab's Submission to L+M Shared Task: All you need is continued pretraining of \chemistry texts even for molecule captioning

workshop paper

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

LLMs have revolutionized the landscape of information retrieval and knowledge dissemination. However, their application in specialized areas is often hindered by limitations such as factual inaccuracies and hallucinations, especially in long-tail knowledge distributions. In this work, we explore the potential of retrieval-augmented generation (RAG) models in performing long-form question answering (LFQA) on a specially curated niche and custom knowledge domain. We present VedantaNY-10M, a dataset curated from extensive public discourses on the ancient Indian philosophy of Advaita Vedanta. We develop and benchmark a RAG model against a standard, non-RAG LLM, focusing on transcription, retrieval, and generation performance. A human evaluation involving computational linguists and domain experts, shows that the RAG model significantly outperforms the standard model in producing factual, comprehensive responses having fewer hallucinations. In addition, we find that a keyword-based hybrid retriever that focuses on unique low-frequency words further improves results. Our study provides insights into the future development of real-world RAG models for custom and niche areas of knowledge.

Ancient Wisdom, Modern Tools: Exploring Retrieval-Augmented LLMs for Ancient Indian Philosophy

Due to ancient origin, there are many incomplete characters in the unearthed Oracle Bone Inscriptions(OBI), which brings the great challenges to recognition and research. In recent years, image inpainting techniques have made remarkable progress. However, these models are unable to adapt to the unique font shape and complex text background of OBI. To meet these aforementioned challenges, we propose a two-stage method for restoring damaged OBI using Generative Adversarial Networks (GAN), which incorporates a dual discriminator structure to capture both global and local image information. In order to accurately restore the image structure and details, the spatial attention mechanism and a novel loss function are proposed. By feeding clear copies of existing OBI and various types of masks into the network, it learns to generate content for the missing regions. Experimental results demonstrate the effectiveness of our proposed method in completing OBI compared to several state-of-the-art techniques.

Coarse-to-Fine Generative Model for Oracle Bone Inscriptions Inpainting

Automatic correction of errors in Handwritten Text Recognition (HTR) output poses persistent challenges yet to be fully resolved. In this study, we introduce a shared task aimed at addressing this challenge, which attracted 271 submissions, yielding only a handful of promising approaches. This paper presents the datasets, the most effective methods, and an experimental analysis in error-correcting HTRed manuscripts and papyri in Byzantine Greek, the language that followed Classical and preceded Modern Greek. By using recognised and transcribed data from seven centuries, the two best-performing methods are compared, one based on a neural encoder-decoder architecture and the other based on engineered linguistic rules. We show that the recognition error rate can be reduced by both, up to 2.5 points at the level of characters and up to 15 at the level of words, while also elucidating their respective strengths and weaknesses.

Challenging Error Correction in Recognised Byzantine Greek

Ancient manuscripts are frequently damaged, containing gaps in the text known as lacunae. In this paper, we present a bidirectional RNN model for character prediction of Coptic characters in manuscript lacunae. Our best model performs with 72% accuracy on single character reconstruction, but falls to 37% when reconstructing lacunae of various lengths. While not suitable for definitive manuscript reconstruction, we argue that our RNN model can help scholars rank the likelihood of textual reconstructions. As evidence, we use our RNN model to rank reconstructions in two early Coptic manuscripts. Our investigation shows that neural models can augment traditional methods of textual restoration, providing scholars with an additional tool to assess lacunae in Coptic manuscripts.

Lacuna Language Learning: Leveraging RNNs for Ranked Text Completion in Digitized Coptic Manuscripts

This work explores the potential of Transformer models focusing on the translation of ancient Egyptian hieroglyphs. We present a novel Hieroglyphic Transformer model, built upon the powerful M2M-100 multilingual translation framework and trained on a dataset we customised from the Thesaurus Linguae Aegyptiae database. Our experiments demonstrate promising results, with the model achieving significant accuracy in translating hieroglyphics into both German and English. This work holds significant implications for Egyptology, potentially accelerating the translation process and unlocking new research approaches.

Deep Learning Meets Egyptology: a Hieroglyphic Transformer for Translating Ancient Egyptian

We present models for lemmatizing and POS-tagging Egyptian, Coptic and Demotic to test the performance of our pipeline for the ancient languages of Egypt. Of these languages, Demotic and Egyptian are known to be difficult to annotate due to their high extent of ambiguity. We report lemmatization accuracy of 86\%, 91\% and 99\%, and XPOS-tagging accuracy of 89\%, 95\% and 98\% for Egyptian, Demotic and Coptic, respectively.

Neural Lemmatization and POS-tagging models for Coptic, Demotic and Earlier Egyptian

Existing Latin treebanks draw from Latin's long written tradition, spanning 17 centuries and a variety of cultures. Recent efforts have begun to harmonize these treebanks' annotations to better train and evaluate morphological taggers. However, the heterogeneity of these treebanks must be carefully considered to build effective and reliable data. In this work, we review existing Latin treebanks to identify the texts they draw from, identify their overlap, and document their coverage across time and genre. We additionally design automated conversions of their morphological feature annotations into the conventions of standard Latin grammar. From this, we build new time-period data splits that draw from the existing treebanks which we use to perform a broad cross-time analysis for POS and morphological feature tagging. We find that BERT-based taggers outperform existing taggers while also being more robust to cross-domain shifts.

Latin Treebanks in Review: An Evaluation of Morphological Tagging Across Time

Hebrew manuscripts preserve thousands of textual transmissions of post-Biblical Hebrew texts from the first millenium. In many cases, the text in the manuscripts is not fully decipherable, whether due to deterioration, perforation, burns, or otherwise. Existing BERT models for Hebrew struggle to fill these gaps, due to the many orthographical deviations found in Hebrew manuscripts. We have pretrained a new dedicated BERT model, dubbed MsBERT (short for: Manuscript BERT), designed from the ground up to handle Hebrew manuscript text. MsBERT substantially outperforms all existing Hebrew BERT models regarding the prediction of missing words in fragmentary Hebrew manuscript transcriptions in multiple genres, as well as regarding the task of differentiating between quoted passages and exegetical elaborations. We provide MsBERT for free download and unrestricted use, and we also provide an interactive and user-friendly website to allow manuscripts scholars to leverage the power of MsBERT in their scholarly work, in order to optimally reconstruct fragmentary Hebrew manuscripts.

MsBERT: A New Model for the Reconstruction of Lacunae in Hebrew Manuscripts

In Japanese, the natural minimal phrase of a sentence is the "bunsetsu" and it serves as a natural boundary of a sentence for native speakers rather than words, and thus grammatical analysis in Japanese linguistics commonly operates on the basis of bunsetsu units.
In contrast, because Japanese does not have delimiters between words, there are two major categories of word definition, namely, Short Unit Words (SUWs) and Long Unit Words (LUWs).
Though a SUW dictionary is available, LUW is not.
Hence, this study focuses on providing deep learning-based (or LLM-based) bunsetsu and Long Unit Words analyzer for the Heian period (AD 794-1185) and evaluating its performances.
We model the parser as transformer-based joint sequential labels model, which combine bunsetsu BI tag, LUW BI tag, and LUW Part-of-Speech (POS) tag for each SUW token.
We train our models on corpora of each period including contemporary and historical Japanese.
The results range from 0.976 to 0.996 in f1 value for both bunsetsu and LUW reconstruction indicating that our models achieve comparable performance with models for a contemporary Japanese corpus.
Through the statistical analysis and diachronic case study, the estimation of bunsetsu could be influenced by the grammaticalization of morphemes.

Long Unit Word Tokenization and Bunsetsu Segmentation of Historical Japanese

Oracle bone script (OBS) is the earliest writing system in China, which is of great value in the improvement of archaeology and Chinese cultural history. However, there are some problems such as the lack of labels and the difficulty to distinguish the glyphs from the background of OBS, which makes the automatic recognition of OBS in the real world not achieve the satisfactory effect. In this paper, we propose a character recognition method based on an unsupervised domain adaptive network (UFCNet). Firstly, a convolutional attention fusion module (CAFM) is designed in the encoder to obtain more global features through multi-layer feature fusion. Second, we construct a Fourier transform (FT) module that focuses on the differences between glyphs and backgrounds. Finally, to further improve the network's ability to recognize character edges, we introduce a kernel norm-constrained loss function. Extensive experiments perform on the Oracle-241 dataset show that the proposed method is superior to other adaptive methods. The code will be available at https://github.com/zhouynan/UFCNet.

Premium content

Knowlab's Submission to L+M Shared Task: All you need is continued pretraining of \chemistry texts even for molecule captioning

Next from ACL 2024

Ancient Wisdom, Modern Tools: Exploring Retrieval-Augmented LLMs for Ancient Indian Philosophy

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES