Large Language Models (LLMs) show remarkable performance on a wide variety of tasks. Most LLMs split text into multi-character tokens and process them as atomic units without direct access to individual characters. This raises the question: To what extent can LLMs learn orthographic information? To answer this, we propose a new benchmark, CUTE, which features a collection of tasks designed to test the orthographic knowledge of LLMs. We evaluate popular LLMs on CUTE, finding that most of them seem to know the spelling of their tokens, yet fail to use this information effectively to manipulate text, calling into question how much of this knowledge is generalizable.

Large Language Models (LLMs) exhibit strong In-Context Learning (ICL) capabilities when prompts with demonstrations are used. However, fine-tuning still remains crucial to further enhance their adaptability. Prompt-based fine-tuning proves to be an effective fine-tuning method in low-data scenarios, but high demands on computing resources limit its practicality. We address this issue by introducing a prompt-based parameter-efficient fine-tuning (PEFT) approach. GNNavi leverages insights into ICL’s information flow dynamics, which indicates that label words act in prompts as anchors for information propagation. GNNavi employs a Graph Neural Network (GNN) layer to precisely guide the aggregation and distribution of information flow during the processing of prompts by hardwiring the desired information flow into the GNN. Our experiments on text classification tasks with GPT-2 and Llama2 show GNNavi surpasses standard prompt-based fine-tuning methods in few-shot settings by updating just 0.2% to 0.5% of parameters. We compare GNNavi with prevalent PEFT approaches, such as prefix tuning, LoRA and Adapter in terms of performance and efficiency. Our analysis reveals that GNNavi enhances information flow and ensures a clear aggregation process.

Prompt-based methods have been successfully applied to multilingual pretrained language models for zero-shot cross-lingual understanding. However, most previous studies primarily focused on sentence-level classification tasks, and only a few considered token-level labeling tasks such as Named Entity Recognition (NER) and Part-of-Speech (POS) tagging. In this paper, we propose Token-Level Prompt Decomposition (ToPro), which facilitates the prompt-based method for token-level sequence labeling tasks. The ToPro method decomposes an input sentence into single tokens and applies one prompt template to each token. Our experiments on multilingual NER and POS tagging datasets demonstrate that ToPro-based fine-tuning outperforms Vanilla fine-tuning and Prompt-Tuning in zero-shot cross-lingual transfer, especially for languages that are typologically different from the source language English. Our method also attains state-of-the-art performance when employed with the mT5 model. Besides, our exploratory study in multilingual large language models shows that ToPro performs much better than the current in-context learning method. Overall, the performance improvements show that ToPro could potentially serve as a novel and simple benchmarking method for sequence labeling tasks.

The NLP community has mainly focused on scaling Large Language Models (LLMs) vertically, i.e., making them better for about 100 languages. We instead scale LLMs horizontally: we create, through continued pretraining, Glot500-m, an LLM that covers 511 predominantly low-resource languages. An important part of this effort is to collect and clean Glot500-c, a corpus that covers these 511 languages and allows us to train Glot500-m. We evaluate Glot500-m on five diverse tasks across these languages. We observe large improvements for both high-resource and low-resource languages compared to an XLM-R baseline. Our analysis shows that no single factor explains the quality of multilingual LLM representations. Rather, a combination of factors determines quality including corpus size, script, “help” from related languages and the total capacity of the model. Our work addresses an important goal of NLP research: we should notlimit NLP to a small fraction of the world’s languages and instead strive to support as many languages as possible to bring the benefits of NLP technology to all languages and cultures. Code, data and models are available at https://github.com/cisnlp/Glot500.

Multilingual Pretrained Language Models (MPLMs) perform strongly in cross-lingual transfer. We propose Prompts Augmented by Retrieval Crosslingually (PARC) to improve zero-shot performance on low-resource languages (LRLs) by augmenting the context with prompts consisting of semantically similar sentences retrieved from a high-resource language (HRL). PARC improves zero-shot performance on three downstream tasks (sentiment classification, topic categorization, natural language inference) with multilingual parallel test sets across 10 LRLs covering 6 language families in unlabeled (+5.1%) and labeled settings (+16.3%). PARC also outperforms finetuning by 3.7%. We find a significant positive correlation between cross-lingual transfer performance on one side, and the similarity between high- and low-resource languages as well as the amount of low-resource pretraining data on the other side. A robustness analysis suggests that PARC has the potential to achieve even stronger performance with more powerful MPLMs.

Pretrained multilingual encoder models can directly perform zero-shot multilingual tasks or linguistic probing by reformulating the input examples into cloze-style prompts. This is accomplished by predicting the probabilities of the label words at the masked token position, without requiring any updates to the model parameters. However, the performance of this method is limited by the model’s bias toward predicting label words which frequently occurred during the pretraining. These words typically receive high probabilities. To address this issue, we combine the models with calibration techniques which modify the probabilities of label words predicted by the models. We first validate the effectiveness of a proposed simple calibration method together with other existing techniques on monolingual encoders in both zero- and few-shot scenarios. We subsequently employ these calibration techniques on multilingual encoders, resulting in substantial performance improvements across a wide range of tasks.

In der Computerlinguistik ist eine kaskadische Prozessierung von Texten üblich. Dabei werden diese zuerst segmentiert (tokenisiert), d.h. Tokens und ggf. Satzgrenzen werden erkannt. Dabei entsteht meist eine Liste bzw. eine einspaltige Tabelle, die sukzessive durch weitere Prozessierungschritte um zusätzliche Spalten – also positionale Annotationen wie z.B. Wortarten und Lemmata für die Tokens in der ersten Spalte – ergänzt wird. Bei der Tokenisierung werden alle Spatien (Leerzeichen) gelöscht. Schon immer problematisch waren dabei Interpunktionszeichen, da diese äußerst ambig sein können, aber auch mehrteilige Namen, die Leerzeichen enthalten und eigentlich zusammengehören. Dieser Beitrag fokussiert auf den Apostroph, der in vielfältiger Weise in den Texten Udo Lindenbergs eingesetzt wird sowie auf mehrteilige Namen, die wir als Tokens erhalten möchten. Wir nutzen dafür das komplette Lindenberg-Archiv des songkorpus.de-Repositoriums, kategorisieren die auftretenden Phänomene, erstellen einen Goldstandard und entwickeln ein teils regel-, teils auf maschinellem Lernen basierendes Segmentierungswerkzeug, das insbesondere die auftretenden Apostrophe, aber auch - lexikonbasiert - mehrteilige Namen nach unseren Vorstellungen erkennt und tokenisiert. Im Anschluss trainieren wir den RNN-Tagger (Schmid, 2019) und zeigen auf, dass ein spezifisch für diese Texte angepasstes Training zu Genauigkeiten ≥ 96% führt. Dabei entsteht nicht nur ein Goldstandard des annotierten Korpus, das dem Songkorpus-Repositorium zur Verfügung gestellt wird, sondern auch eine angepasste Version des RNN-Taggers (verfügbar auf github), die für ähnliche Texte verwendet werden kann.

Constituency parsing plays a fundamental role in advancing natural language processing (NLP) tasks. However, training an automatic syntactic analysis system for ancient languages solely relying on annotated parse data is a formidable task due to the inherent challenges in building treebanks for such languages. It demands extensive linguistic expertise, leading to a scarcity of available resources. To overcome this hurdle, cross-lingual transfer techniques which require minimal or even no annotated data for low-resource target languages offer a promising solution. In this study, we focus on building a constituency parser for Middle High German (MHG) under realistic conditions, where no annotated MHG treebank is available for training. In our approach, we leverage the linguistic continuity and structural similarity between MHG and Modern German (MG), along with the abundance of MG treebank resources. Specifically, by employing the delexicalization method, we train a constituency parser on MG parse datasets and perform cross-lingual transfer to MHG parsing. Our delexicalized constituency parser demonstrates remarkable performance on the MHG test set, achieving an F1-score of 67.3%. It outperforms the best zero-shot cross-lingual baseline by a margin of 28.6% points. The encouraging results underscore the practicality and potential for automatic syntactic analysis in other ancient languages that face similar challenges as MHG.

We present a literature and empirical survey that critically assesses the state of the art in character-level modeling for machine translation (MT). Despite evidence in the literature that character-level systems are comparable with subword systems, they are virtually never used in competitive setups in WMT competitions. We empirically show that even with recent modeling innovations in character-level natural language processing, character-level MT systems still struggle to match their subword-based counterparts. Character-level MT systems show neither better domain robustness, nor better morphological generalization, despite being often so motivated. However, we are able to show robustness towards source side noise and that translation quality does not degrade with increasing beam size at decoding time.

A recent approach for few-shot text classification is to convert textual inputs to cloze questions that contain some form of task description, process them with a pretrained language model and map the predicted words to labels. Manually defining this mapping between words and labels requires both domain expertise and an understanding of the language model’s abilities. To mitigate this issue, we devise an approach that automatically finds such a mapping given small amounts of training data. For a number of tasks, the mapping found by our approach performs almost as well as hand-crafted label-to-word mappings.

We present our systems for the WMT20 Very Low Resource MT Task for translation between German and Upper Sorbian. For training our systems, we generate synthetic data by both back- and forward-translation. Additionally, we enrich the training data with German-Czech translated from Czech to Upper Sorbian by an unsupervised statistical MT system incorporating orthographically similar word pairs and transliterations of OOV words. Our best translation system between German and Sorbian is based on transfer learning from a Czech-German system and scores 12 to 13 BLEU higher than a baseline system built using the available parallel data only.

We present a generative model that efficiently mines transliteration pairs in a consistent fashion in three different settings: unsupervised, semi-supervised, and supervised transliteration mining. The model interpolates two sub-models, one for the generation of transliteration pairs and one for the generation of non-transliteration pairs (i.e., noise). The model is trained on noisy unlabeled data using the EM algorithm. During training the transliteration sub-model learns to generate transliteration pairs and the fixed non-transliteration model generates the noise pairs. After training, the unlabeled data is disambiguated based on the posterior probabilities of the two sub-models. We evaluate our transliteration mining system on data from a transliteration mining shared task and on parallel corpora. For three out of four language pairs, our system outperforms all semi-supervised and supervised systems that participated in the NEWS 2010 shared task. On word pairs extracted from parallel corpora with fewer than 2% transliteration pairs, our system achieves up to 86.7% F-measure with 77.9% precision and 97.8% recall.

This paper describes general requirements for evaluating and documenting NLP tools with a focus on morphological analysers and the design of a Gold Standard. It is argued that any evaluation must be measurable and documentation thereof must be made accessible for any user of the tool. The documentation must be of a kind that it enables the user to compare different tools offering the same service, hence the descriptions must contain measurable values. A Gold Standard presents a vital part of any measurable evaluation process, therefore, the corpus-based design of a Gold Standard, its creation and problems that occur are reported upon here. Our project concentrates on SMOR, a morphological analyser for German that is to be offered as a web-service. We not only utilize this analyser for designing the Gold Standard, but also evaluate the tool itself at the same time. Note that the project is ongoing, therefore, we cannot present final results.

In the framework of the preparation of linguistic web services for corpus processing, the need for a representation format was felt, which supports interoperability between different web services in a corpus processing pipeline, but also provides a well-defined interface to both, legacy tools and their data formats and upcoming international standards. We present the D-SPIN text corpus format, TCF, which was designed for this purpose. It is a stand-off XML format, inspired by the philosophy of the emerging standards LAF (Linguistic Annotation Framework) and its ``instances'' MAF for morpho-syntactic annotation and SynAF for syntactic annotation. Tools for the exchange with existing (best practice) formats are available, and a converter from MAF to TCF is being tested in spring 2010. We describe the usage scenario where TCF is embedded and the properties and architecture of TCF. We also give examples of TCF encoded data and describe the aspects of syntactic and semantic interoperability already addressed.

It is proposed to parse feature structure-based grammars in several steps. Each step is aimed to eliminate as many invalid analyses as possible as efficiently as possible. To this end the set of feature constraints is divided into three subsets, a set of context-free constraints, a set of filtering constraints and a set of structure-building constraints, which are solved in that order. The best processing strategy differs: Context-free constraints are solved efficiently with one of the well-known algorithms for context-free parsing. Filtering constraints can be solved using unification algorithms for non-disjunctive feature structures whereas structure-building constraints require special techniques to represent feature structures with embedded disjunctions efficiently. A compilation method and an efficient processing strategy for filtering constraints are presented.