The most common way to listen recorded music nowadays is via streaming platforms which provide access tens of millions tracks. To assist users in effectively browsing these large catalogs, the integration Music Recommender Systems (MRSs) has become essential. Current real-world MRSs are often quite complex and optimized for recommendation accuracy. They combine several building blocks based on collaborative filtering content-based recommendation. This complexity can hinder ability explain...

Process mining has been successfully used in automatic knowledge discovery and providing guidance or support. The known process approaches rely on processes being executed with the help of information systems thus enabling capture traces as event logs. However, there are many other fields such Humanities, Social Sciences Medicine where workers follow log their execution manually textual forms instead. problem we tackle this paper is instance models from unstructured, text-based traces. Using...

News organizations employ personalized recommenders to target news articles specific readers and thus foster engagement. Existing approaches rely on extensive user profiles. However frequently possible, rarely authenticate themselves publishers' websites. This paper proposes an approach for such cases. It provides a basic degree of personalization while complying with the key characteristics recommendation including popularity, recency, dynamics reading behavior. We extend existing research...

Music listening context such as location or activity has been shown to greatly influence the users' musical tastes. In this work, we study relationship between user and audio content in order enable context-aware music recommendation agnostic data. For that, propose a semi-automatic procedure collect track sets which leverages playlist titles proxy for labelling. Using this, create release dataset of ~50k tracks labelled with 15 different contexts. Then, present benchmark classification...

The problem of multi-label classification with missing labels (MLML) is a common challenge that prevalent in several domains, e.g. image annotation and auto-tagging. In classification, each instance may belong to multiple class simultaneously. Due the nature dataset collection labelling procedure, it have incomplete annotations dataset, i.e. not all samples are labelled corresponding labels. However, data hinders training models. MLML has received much attention from research community....

The strategic transition of media organizations to personalized information delivery has urged the need for richer methods analyze customers. Though useful in supporting creation recommender strategies, current data mining techniques create complex models requiring often an understanding order interpret results. This situation together with technologies deluge and particularities news industry pose challenges organization making decisions about most suitable strategy. Therefore, we propose...

We conducted a human subject study of named entity recognition on noisy corpus conversational music recommendation queries, with many irregular and novel entities. evaluated the NER linguistic behaviour in these challenging conditions compared it most common systems nowadays, fine-tuned transformers. Our goal was to learn about task guide design better evaluation methods algorithms. The results showed that our context quite hard for both algorithms under strict schema; humans had higher...

We conducted a human subject study of named entity recognition on noisy corpus conversational music recommendation queries, with many irregular and novel entities. evaluated the NER linguistic behaviour in these challenging conditions compared it most common systems nowadays, fine-tuned transformers. Our goal was to learn about task guide design better evaluation methods algorithms. The results showed that our context quite hard for both algorithms under strict schema; humans had higher...

Despite impressive results of language models for named entity recognition (NER), their generalization to varied textual genres, a growing type set, and new entities remains challenge. Collecting thousands annotations in each case training or fine-tuning is expensive time-consuming. In contrast, humans can easily identify given some simple instructions. Inspired by this, we challenge the reliance on large datasets study pre-trained NER meta-learning setup. First, test typing (NET) zero-shot...

Annotating music items with genres is crucial for recommendation and information retrieval, yet challenging given that are subjective concepts. Recently, in order to explicitly consider this subjectivity, the annotation of was modeled as a translation task: predict item its within target vocabulary or taxonomy (tag system) from set genre tags originating other tag systems. However, without parallel corpus, previous solutions could not handle systems languages, being limited English-language...

In recent years, generated content in music has gained significant popularity, with large language models being effectively utilized to produce human-like lyrics various styles, themes, and linguistic structures. This technological advancement supports artists their creative processes but also raises issues of authorship infringement, consumer satisfaction spamming. To address these challenges, methods for detecting are necessary. However, existing works have not yet focused on this specific...

Large Language Models (LLMs) zero-shot and few-shot performance are subject to memorization data contamination, complicating the assessment of their validity. In literary tasks, LLMs is often correlated degree book memorization. this work, we carry out a realistic evaluation for quotation attribution in novels, taking instruction fined-tuned version Llama3 as an example. We design task-specific measure use it show that Llama3's ability perform positively novel However, still performs...

Humans naturally attribute utterances of direct speech to their speaker in literary works. When attributing quotes, we process contextual information but also access mental representations characters that build and revise throughout the narrative. Recent methods automatically such have explored simulating human logic with deterministic rules or learning new implicit neural networks when processing information. However, these systems inherently lack \textit{character} representations, which...

Recommender systems relying on Language Models (LMs) have gained popularity in assisting users to navigate large catalogs. LMs often exploit item high-level descriptors, i.e. categories or consumption contexts, from training data user preferences. This has been proven effective domains like movies products. However, the music domain, understanding how effectively utilize song descriptors for natural language-based recommendation is relatively limited. In this paper, we assess effectiveness...

Online news readers exhibit a very dynamic behavior. News publishers have been investigating ways to predict such changes in order adjust their recommendation strategies and better engage the readers. Existing research focuses on analyzing evolution of reading interests associated with categories. Compared these, we study also how relations among change time. Observations over 10-month period German publisher indicate that overall, amid categories change, but stable periods spanning months...

Prevalent efforts have been put in automatically inferring genres of musical items. Yet, the propose solutions often rely on simplifications and fail to address diversity subjectivity music genres. Accounting for these has, though, many benefits aligning knowledge sources, integrating data enriching items with tags. Here, we choose a new angle genre study by seeking predict what would be target tag system, knowing assigned them within source systems. We call this translation task identify...