Question:

Are individual allowed to use accrual based accounting for federal income tax? · 2025

Read Question: by Are individual allowed to use accrual based accounting for federal income tax? (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-00725.txt.

Introduction

Bithiah Yuan builds a finance question-answering system in a 2025 paper. Advisers face floods of mixed text and tables, and a system that picks the right passage as an answer could speed their calls. Finance text is scarce and domain-bound, which starves standard models.

The system pairs a fast keyword retriever with a re-ranker built from pre-trained language models. Training tricks adapt the model to finance phrasing, and a pointwise scheme scores each candidate. On a public finance benchmark's hard task, the best setup beat prior records on all three ranking measures by wide margins. Transfer and adapt, the paper shows, is the winning recipe.

Catalog record

CREATORAre individual allowed to use accrual based accounting for federal income tax?
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE84 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-00725

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Question: Are individual allowed to use accrual based accounting for federal income tax? Abstract Master’s Thesis \thetitle \theauthor Examiner: Dr. Fang Wei-Kleiner Adviser: Dr. Fang Wei-Kleiner Albert-Ludwigs-University Freiburg Faculty of Engineering Department of Computer Science Chair of Databases and Information Systems July 31, 2020 Writing Period 31. 01. 2020 – 31. 07. 2020 Examiner Dr. Fang Wei-Kleiner Second Examiner Prof. Dr. Josif Grabocka Adviser Dr. Fang Wei-Kleiner Declaration I hereby declare, that I am the sole author and composer of my thesis and that no other sources or learning aids, other than those listed, have been used. Furthermore, I declare that I have acknowledged the work of others by providing detailed references of said work. I also hereby declare that my thesis has not been prepared for another examination or assignment, either in its entirety or excerpts thereof. Place, Date Signature Abstract Motivated by the emerging demand in the financial industry for the automatic analysis of unstructured and structured data at scale, Question Answering (QA) systems can provide lucrative and competitive advantages to companies by facilitating the decision making of financial advisers. Consequently, we propose a novel financial QA system using the Transformer-based pre-trained BERT language model to solve the limitations of data scarcity and language specificity in the financial domain. Our system focuses on financial non-factoid answer selection, which retrieves a set of passage-level texts and selects the most relevant as the answer. To increase the efficiency, we formulate the answer selection task as a re-ranking problem, in which our system consists of an Answer Retriever using BM25, a simple information retrieval approach, to first return a list of candidate answers and an Answer Re-ranker built with variants of pre-trained BERT language models to re-rank and select the most relevant answers. We investigate various learning, further pre-training, and fine-tuning approaches for BERT and our experiments suggest that FinBERT-QA, a model built from applying the Transfer and Adapt further fine-tuning and pointwise learning approach is the most effective as it improves the state-of-the-art results of task 2 of the FiQA dataset by 16% on MRR, 17% on NDCG, and 21% on Precision@1. Chapter 1 Introduction In recent years, Natural Language Processing (NLP) techniques have proven to be powerful due to their ability to automatically analyze domain-specific data at scale. The financial domain, in particular, relies on translating multiple unstructured and structured data for fast and detailed decision making [ 5 ] . The use of sentiment analysis in the financial domain to predict the behavior of stock markets or credit risks have been widely researched and shown to be effective. This motivates further research of applying NLP methods such as Question Answering (QA) to the financial domain. QA systems can be used to assist financial advisers of large institutions answer financial outlook and information questions from their clients. In order to answer questions that examine how financial entities, such as company stocks, currencies, industries, and assets are expected to perform in the future, the advisers need to provide textual passages with answers based on past or recent financial reports. However, retrieving financial reports pose a challenge since the experience levels of financial advisers vary, which results in inconsistent interpretations of the reports. Moreover, the large magnitude of detailed and technically written incoming financial reports are infeasible to read through even for the best and most experienced advisers [ 5 ] . As a result, the application of QA in the financial domain is critical due to the highly competitive and profitable nature of the industry. Since QA research targeting financial datasets is underexplored and large profits and competitive advantages can be achieved with slight improvements in the underlying technology [ 6 ] , further experimentation using NLP and Deep Learning approaches can greatly benefit financial industries in the current era of digital transformation. Following the need for a novel financial QA system, Section 1 further reinforces our motivation, Section 2 provides a formal task definition, and Section 3 provides an overview of our approach and contributions. 1 Motivation Financial QA is more of a challenge compared to general QA due to the specialized terminology involved. For example, a sentence such as "there is an increase in the short sells of a company" requires domain-specific knowledge to know that it implies a negative outlook [ 5 ] . Early NLP approaches for financial QA are based on feature engineering financial lexicons, for example, using handcrafted features based on existing financial knowledge to build a classifier predicting if an answer is relevant to a question [ 5 ] . However, feature engineering and simple word counting are not enough since the correct answer may only be semantically related to the question. Moreover, the questions and answers may not share direct lexical units. For example, the correct answer could be formulated differently or serve to complete the question with missing information [ 28 ] . The correct answers can also be noisy and contain large amounts of irrelevant information and words that are not in the question [ 27 ] . As depicted in Table 1 , which is an example of a question, ground truth answer, and an incorrect negative answer from the FiQA dataset [ 15 ] , QA in the financial domain is especially difficult because even though the negative answer has shared lexical units with the question such as "income" and "accrual", it does not answer the question semantically. Deep learning models using neural networks have exploded in recent years and have shown to be effective as they require little feature engineering and are able to better learn the context in a language corpus. However, traditional deep learning methods based on representation learning have the following drawbacks when applied to financial texts: 1. Data Scarcity: Deep learning models require a large amount of training data to understand the underlying latent patterns of the corpora. This presents a problem because large labeled financial datasets are scarce and building such a large-scale dataset would be complex and expensive since it requires annotations from highly trained financial experts [ 2 ] . 2. Financial Language Specificity: Financial corpora have specific language and underlying conceptualizations. Therefore, only training a model on the target corpus can fail to capture particular semantic and contextual meanings [ 2 ] . 3. Shallow pre-trained Word Embeddings Usage: Since many deep learning approaches are based on shallow pre-trained word embedding such as word2vec [ 19 ] or GloVe [ 22 ] , they are not able to deeply capture the contextual meanings of financial datasets [ 2 ] . Recently, there have been large developments in using transfer learning methods for NLP tasks in order to solve the above challenges. In particular, pre-trained language models such as ELMo [ 23 ] , OpenAI GPT [ 24 ] and BERT [ 7 ] have achieved several breakthroughs in the last two years. Pre-trained language models are extremely effective since they are first pre-trained on a large-scale corpora in an unsupervised manner, then the entire pre-trained model is applied to a downstream task to better capture the context and dependencies between words. The Transformer-based pre-trained language model, BERT, in particular, can be easily applied to different NLP tasks even if

Read the complete text