1,721,077 research outputs found
Beyond traditional search: bridging retrieval, reasoning, and language barriers in intelligent search systems
Get PDFThis thesis investigates innovative approaches to address fundamental limitations in modern search systems, with a particular focus on enhancing the synergy between retrieval and reasoning components and improving the ease of information access. A primary research thread explores the optimization of interactions between retrievers and reasoning components. We specifically address the challenge of "hard false relevant" documents—those that appear superficially relevant but lack true semantic alignment with queries. To tackle this issue, we introduce Reinforced Retrieval Augmented Machine Learning (RRAML), a visionary framework that enables retrieval systems to be fine-tuned within a Retrieval-Augmented Generation (RAG) architecture. This novel approach allows the retriever to adapt through continued training when combined with a reasoner in a RAG scenario, to reduce the retrieval of false relevant documents.
Within the same research direction, we explored Neural Semantic Parsing (NSP), which uses Large Language Models (LLMs) to translate natural language queries into a machine-readable format that can be used to retrieve information from knowledge graphs. This approach represents an alternative to traditional RAG systems, where, instead of retrieving from unstructured text, the LLM facilitates access to structured and verified information stored in knowledge graphs, providing greater control and transparency in the information retrieval process. To enhance this system's reliability, we developed the Hallucination Simulation Framework, which deliberately induces hallucinations in semantic parsers during training. Complementing this, we created the Hallucination Detection Model (HDM), which identifies and mitigates hallucinations stemming from knowledge gaps, improving answer reliability by 20\%. This framework enables semantic parsers to recognize their knowledge boundaries and uncertainty levels, resulting in more transparent and trustworthy responses.
To further support the democratization of information, this thesis introduces approaches to bridge linguistic and cultural barriers, enabling users worldwide to access accurate, relevant information regardless of language or technical proficiency. This goal led to the development of a new architecture, x-NDB, for cross-lingual neural databases and the creation of X-WikiNLDB, a dataset containing unstructured text in multiple languages simulating data retrieved online. X-WikiNLDB facilitates robust cross-lingual information retrieval. Our cross-lingual performance are on-par with previous work in the English-only scenario. Furthermore, we demonstrated significant zero-shot performance improvements of 2-5 compared to the multilingual counterpart across several low-resource languages in Catalan, Tagalog, Yoruba, Japanese, and Korean. This success suggests that cross-lingual training encourages models to capture deeper semantic understanding rather than surface-level patterns, enabling better generalization across linguistic boundaries.
To advance language accessibility further, we developed Fauno, DanteLLM, and OpenDanteLLM, pioneering a series of open-source Italian language models. Starting with Fauno, we created the first open-source conversational Italian LLM along with a novel conversational dataset. Building upon this foundation, DanteLLM achieved remarkable performance improvements, demonstrating a 10\% increase over Fauno and 7\% over the best competitor across comprehensive Italian benchmarks. OpenDanteLLM, while trained exclusively on open-source data, showed a 5\% improvement over existing methods while ensuring unrestricted access through its commercial-friendly Apache 2.0 license. This work, which gained particular relevance during ChatGPT's temporary ban in Italy, has established a new direction in language-specific AI development, inspiring other researchers to create additional Italian language models. Our approach demonstrates the feasibility of building high-performance, privacy-preserving language models that operate entirely offline, providing Italian-speaking users with reliable alternatives to centralized systems.
The thesis also addresses the challenge of ambiguous queries and the dependence on behavioral data in search systems. We introduce Generative Query Recommendation (GQR), a zero-shot approach that reimagines query expansion without relying on user query logs. By leveraging LLMs as the sole component, GQR eliminates the complex pipelines and query log dependencies associated with traditional methods. Our system significantly outperformed industry standards, demonstrating a 10-point improvement in NDCG@10. The generated query reformulations showed reduced ambiguity with respect to the document collection, as measured by a 7-point increase in the Simplified Clarity Score. A blind user study with 12 annotators further validated GQR's effectiveness, with users preferring our recommendations approximately 60\% of the time compared to leading industry alternatives. This approach establishes a new paradigm for query recommendations that achieves superior retrieval accuracy and user satisfaction without requiring behavioral data.
Finally, we present the Multi-Relevant Future Items Evaluation (MRFI) protocol for Sequential Recommender Systems (SRS), which improves evaluation by considering multiple relevant items. MRFI, alongside a novel loss function that integrates relevance feedback, enhances recommendation accuracy and reliability, achieving improvements of 2.82 points of NDCG@10 and 0.64\% Hit Rate across several benchmark datasets. This methodological innovation provides a robust basis for evaluating and training sequential recommendation systems across diverse applications.
Together, these contributions address the core challenges of relevance, reliability, and accessibility in information retrieval and language technology, supporting the vision of universally accessible, high-quality information systems
Long-term results of external auditory canal closure and mastoid obliteration in cochlear implantation after radical mastoidectomy: a clinical and radiological study
No full textAbstract Cochlear implant candidates with mastoid cavity present a significant challenge to safe cochlear implantation because of possible spread of infection to the inner ear as well as an increased risk of electrode array extrusion. Closure of the external auditory canal is one of the several surgical techniques utilized to block the potential entry routes for infection and to protect the implanted device. The main concern after external auditory canal closure is the risk of developing a cholesteatoma, which can lead to an asymptomatic erosion of the temporal bone and/or cochlear implant failure. In this study we present the results of very long-term (mean 12 years) clinical and radiological follow-up in 12 patients who underwent external auditory canal closure associated with mastoid and Eustachian tube obliteration to facilitate cochlear implantation. To date, with a mean ± SD follow- up of 12 ± 4.7 years (range 5–21 years), the only com- plication experienced was the breakdown of the EAC closure in one patient, successfully treated by performing a rotation skin flap. The results of this study confirmed that external auditory canal closure is a reliable technique in cochlear implantation after radical mastoidectomy pro- vided that a rigorous surgical technique is performed. A right balance between the need to reduce costs and to avoid unnecessary doses of radiation to patients and the task of a radiological surveillance may be represented by perform- ing computed tomography 12–18 months postoperatively and then, only if clinically warranted
The pectoralis major myocutaneous flap in head and neck reconstructiv surgery: 16 years of experience
No full textMiddle ear and mastoid glomus tumors (glomus tympanicum): An algorithm for the surgical management.
No full textAuris Nasus Larynx. 2010 Dec;37(6):661-8. Epub 2010 Apr 18.
Middle ear and mastoid glomus tumors (glomus tympanicum): an algorithm for the surgical management.
Sanna M, Fois P, Pasanisi E, Russo A, Bacciu A.
SourceGruppo Otologico, Piacenza-Rome, Italy.
Abstract
OBJECTIVE: Glomus tumors are slow-growing benign lesions and represent the most common primary neoplasms of the middle ear. The objective of the present study is to report our surgical strategy in the management of glomus tympanicum tumors.
METHODS: Between December 1988 and July 2008, 68 patients with histologically confirmed glomus tympanicum tumor underwent surgical treatment. The follow-up of the series ranged from 6 to 194 months (mean, 33.4 months).
RESULTS: Distribution of tumors according to Fisch and Mattox classification was as follows: type A, 44 cases (64.7%); type B, 24 cases (35.3%). All of the 44 Class A tumors were safely removed via either a stapedectomy-type transcanal approach or a retroauricolar-transcanal approach. Five patients with Class B tumors were operated on through a transmastoid approach. Nineteen patients with larger Class B tumors underwent a subtotal petrosectomy with blind sac closure of the external auditory canal and middle ear obliteration. Gross total tumor removal was achieved in all 68 cases. In one case (1.4%) there was recurrence after 9 years, for which the patient was re-operated. No residual/recurrence has been detected on computed tomography in the rest of the patients.
CONCLUSION: Surgical excision of glomus tympanicum tumors is a safe procedure, allowing total tumor removal with minimal morbidity, a low recurrence rate, and a low complication rate
Cochlear implantation in chronic otitis media and previous middle ear surgery: 20 years of experience
No full textCochlear implantation in the setting of chronic otitis media or previous middle ear surgery poses several problems for the surgeon: possible spread of infection to the cochlea and the subarachnoid spaces with consequent meningitis, risk of electrode array extrusion and possible recurrence of the original disease. Several surgical strategies have been proposed to overcome these problems. In the present study, clinical and functional results of cochlear implantation in 26 patients with chronic otitis media (8 cases) or previous middle ear surgery (18 cases) in the ear most suitable for implantation were retrospectively reviewed. Among the 8 patients with chronic otitis media, in 7 cases a subtotal petrosectomy associated with external auditory canal closure and mastoid and Eustachian tube obliteration was performed, while in the remaining patient cochlear implantation was done 6 months after a myringoplasty. The only complication observed was a reperforation of the tympanic membrane in this latter patient. Among the 18 patients with previous middle ear surgery, 2 had undergone intact canal wall tympanomastoidectomy and were implanted utilising the previous surgical approach. In the remaining 16 patients who had a radical cavity, an open technique was maintained in 3 cases; a cavity revision associated to external auditory canal closure, Eustachian tube and mastoid obliteration was performed in 12 patients, while in one case a middle cranial fossa approach was utilised. Two of the 3 patients in whom an open technique was maintained have experienced electrode array extrusion. The only complication observed in the remaining patients was the breakdown of the external auditory canal closure in one case. No problems were noted in patients who had undergone intact canal wall tympanomastoidectomy as well as in the subject implanted via the middle cranial fossa approach. All patients achieved and maintained good hearing performance over time. Subtotal petrosectomy associated with external auditory canal closure, Eustachian tube occlusion and mastoid obliteration is an effective procedure to facilitate cochlear implantation in presence of chronic otitis media. The open cavity technique offers the advantage of a close clinical examination, but may expose the patient to the risk of electrode array extrusion, mainly in the long-term period
Bilateral vocal fold immobility in a patient with overlap syndrome rheumatoid arthritis/systemic sclerosis.
No full textLembo miocutaneo di grande pettorale nella chirurgia riparativa cervico-facciale.Nostra esperienza.
No full textSurgical management of jugular foramen schwannomas with hearing and facial nerve function preservation: a series of 23 cases and review of the literature.
No full text
- …
