Preventing student dropout is a major challenge in higher education and it is difficult to predict prior to enrolment which students are likely to drop out and which students are likely to succeed. High School GPA is a strong predictor of dropout, but much variance in dropout remains to be explained. This study focused on predicting university dropout by using text mining techniques with the aim of exhuming information contained in motivation statements written by students. By combining text data with classic predictors of dropout in the form of student characteristics, we attempt to enhance the available set of predictive student characteristics. Our dataset consisted of 7,060 motivation statements of students enrolling in a non-selective bachelor at a Dutch university in 2014 and 2015. Support Vector Machines were trained on 75 percent of the data and several models were estimated on the test data. We used various combinations of student characteristics and text, such as TFiDF, topic modelling, LIWC dictionary. Results showed that, although the combination of text and student characteristics did not improve the prediction of dropout, text analysis alone predicted dropout similarly well as a set of student characteristics. Suggestions for future research are provided.

Existing agents for solving tasks such as ML engineering rely on prompting powerful language models. As a result, these agents do not improve with more experience. In this paper, we show that agents backed by weaker models that improve via reinforcement learning (RL) can outperform agents backed by much larger, but static models. We identify two major challenges with RL in this setting. First, actions can take a variable amount of time (e.g., executing code for different solutions), which leads to asynchronous policy gradient updates that favor faster but suboptimal solutions. To tackle variable-duration actions, we propose duration-aware gradient updates in a distributed asynchronous RL framework to amplify high-cost but high-reward actions. Second, using only test split performance as a reward provides limited feedback. A program that is nearly correct is treated the same as one that fails entirely. To address this, we propose environment instrumentation to offer partial credit, distinguishing almost-correct programs from those that fail early (e.g., during data loading). Environment instrumentation uses a separate static language model to insert print statement to an existing program to log the agent's experimental progress, from which partial credit can be extracted as reward signals for learning. Our experimental results on MLEBench suggest that performing gradient updates on a much smaller model (Qwen2.5-3B) trained with RL outperforms prompting a much larger model (Claude-3.5-Sonnet) with agent scaffolds, by an average of 22% across 12 Kaggle tasks.

Central bank communication plays a critical role in shaping economic expectations and monetary policy effectiveness. This study applies supervised machine learning techniques to classify the sentiment of press releases from the Bank of Thailand, addressing gaps in research that primarily focus on lexicon-based approaches. My findings show that supervised learning can be an effective method, even with smaller datasets, and serves as a starting point for further automation. However, achieving higher accuracy and better generalization requires a substantial amount of labeled data, which is time-consuming and demands expertise. Using models such as Na\"ive Bayes, Random Forest and SVM, this study demonstrates the applicability of machine learning for central bank sentiment analysis, with English-language communications from the Thai Central Bank as a case study.

The inaugural ACM International Conference on AI-powered Software introduced the AIware Challenge, prompting researchers to explore AI-driven tools for optimizing conference programs through constrained optimization. We investigate the use of Large Language Models (LLMs) for program scheduling, focusing on zero-shot learning and integer programming to measure paper similarity. Our study reveals that LLMs, even under zero-shot settings, create reasonably good first drafts of conference schedules. When clustering papers, using only titles as LLM inputs produces results closer to human categorization than using titles and abstracts with TFIDF. The code has been made publicly available.

Today text classification becomes critical task for concerned individuals for numerous purposes. Hence, several researches have been conducted to develop automatic text classification for national and international languages. However, the need for an automatic text categorization system for local languages is felt. The main aim of this study is to establish a Pashto automatic text classification system. In order to pursue this work, we built a Pashto corpus which is a collection of Pashto documents due to the unavailability of public datasets of Pashto text documents. Besides, this study compares several models containing both statistical and neural network machine learning techniques including Multilayer Perceptron (MLP), Support Vector Machine (SVM), K Nearest Neighbor (KNN), decision tree, gaussian na\"ive Bayes, multinomial na\"ive Bayes, random forest, and logistic regression to discover the most effective approach. Moreover, this investigation evaluates two different feature extraction methods including unigram, and Time Frequency Inverse Document Frequency (IFIDF). Subsequently, this research obtained average testing accuracy rate 94% using MLP classification algorithm and TFIDF feature extraction method in this context.

Over the last few years, neural network derived word embeddings became popular in the natural language processing literature. Studies conducted have mostly focused on the quality and application of word embeddings trained on public available corpuses such as Wikipedia or other news and social media sources. However, these studies are limited to generic text and thus lack technical and scientific nuances such as domain specific vocabulary, abbreviations, or scientific formulas which are commonly used in academic context. This research focuses on the performance of word embeddings applied to a large scale academic corpus. More specifically, we compare quality and efficiency of trained word embeddings to TFIDF representations in modeling content of scientific articles. We use a word2vec skip-gram model trained on titles and abstracts of about 70 million scientific articles. Furthermore, we have developed a benchmark to evaluate content models in a scientific context. The benchmark is based on a categorization task that matches articles to journals for about 1.3 million articles published in 2017. Our results show that content models based on word embeddings are better for titles (short text) while TFIDF works better for abstracts (longer text). However, the slight improvement of TFIDF for larger text comes at the expense of 3.7 times more memory requirement as well as up to 184 times higher computation times which may make it inefficient for online applications. In addition, we have created a 2-dimensional visualization of the journals modeled via embeddings to qualitatively inspect embedding model. This graph shows useful insights and can be used to find competitive journals or gaps to propose new journals.

Term weighting schemes are widely used in Natural Language Processing and Information Retrieval. In particular, term weighting is the basis for keyword extraction. However, there are relatively few evaluation studies that shed light about the strengths and shortcomings of each weighting scheme. In fact, in most cases researchers and practitioners resort to the well-known tf-idf as default, despite the existence of other suitable alternatives, including graph-based models. In this paper, we perform an exhaustive and large-scale empirical comparison of both statistical and graph-based term weighting methods in the context of keyword extraction. Our analysis reveals some interesting findings such as the advantages of the less-known lexical specificity with respect to tf-idf, or the qualitative differences between statistical and graph-based methods. Finally, based on our findings we discuss and devise some suggestions for practitioners. We release our code at https://github.com/asahi417/kex .

Finding an expert plays a crucial role in driving successful collaborations and speeding up high-quality research development and innovations. However, the rapid growth of scientific publications and digital expertise data makes identifying the right experts a challenging problem. Existing approaches for finding experts given a topic can be categorised into information retrieval techniques based on vector space models, document language models, and graph-based models. In this paper, we propose $\textit{ExpFinder}$, a new ensemble model for expert finding, that integrates a novel $N$-gram vector space model, denoted as $n$VSM, and a graph-based model, denoted as $\textit{$\mu$CO-HITS}$, that is a proposed variation of the CO-HITS algorithm. The key of $n$VSM is to exploit recent inverse document frequency weighting method for $N$-gram words and $\textit{ExpFinder}$ incorporates $n$VSM into $\textit{$\mu$CO-HITS}$ to achieve expert finding. We comprehensively evaluate $\textit{ExpFinder}$ on four different datasets from the academic domains in comparison with six different expert finding models. The evaluation results show that $\textit{ExpFinder}$ is a highly effective model for expert finding, substantially outperforming all the compared models in 19% to 160.2%.

Although machine learning has become a powerful tool to augment doctors in clinical analysis, the immense amount of labeled data that is necessary to train supervised learning approaches burdens each development task as time and resource intensive. The vast majority of dense clinical information is stored in written reports, detailing pertinent patient information. The challenge with utilizing natural language data for standard model development is due to the complex nature of the modality. In this research, a model pipeline was developed to utilize an unsupervised approach to train an encoder-language model, a recurrent network, to generate document encodings; which then can be used as features passed into a decoder-classifier model that requires magnitudes less labeled data than previous approaches to differentiate between fine-grained disease classes accurately. The language model was trained on unlabeled radiology reports from the Massachusetts General Hospital Radiology Department (n=218,159) and terminated with a loss of 1.62. The classification models were trained on three labeled datasets of head CT studies of reported patients, presenting large vessel occlusion (n=1403), acute ischemic strokes (n=331), and intracranial hemorrhage (n=4350), to identify a variety of different findings directly from the radiology report data; resulting in AUCs of 0.98, 0.95, and 0.99, respectively, for the large vessel occlusion, acute ischemic stroke, and intracranial hemorrhage datasets. The output encodings are able to be used in conjunction with imaging data, to create models that can process a multitude of different modalities. The ability to automatically extract relevant features from textual data allows for faster model development and integration of textual modality, overall, allowing clinical reports to become a more viable input for more encompassing and accurate deep learning models.

We plot the judge's probability of the target answer given that sentence against how often humans also select that target answer given that same sentence. Humans tend to find a sentence to be strong evidence for an answer when the judge model finds it to be strong evidence. Strong evidence to a model tends to be strong evidence to humans as shown in Figure 7. Combined with the previous result, we can see that learned agents are more accurate at predicting sentences that humans find to be strong evidence. F Model Evaluation of Evidence on DREAM Figure 8 shows how convincing various judge models find each evidence agent. Our findings on DREAM are similar to those from RACE in §4.2. Figure 8: On DREAM, how often each judge selects an agent's answer when given a single agent-chosen sentence. The black line divides learned agents (right) and search agents (left), with human evidence selection in the leftmost column. All agents find evidence that convinces judge models more often than a no-evidence baseline (33%). Learned agents predicting p ( i) or p ( i) find the most broadly convincing evidence.