Shuyi

[Shuyi @Shuyi817]

With just a simple, but powerful calculation like the Cosine Similarity, we can build a reasonable recommendation system.
#aigraph
https://t.co/ROEAamxno2

[Shuyi @Shuyi817]

LLM is easily distracted by irrelevant information, but we can mitigate it by simply adding an instruction to ignore irrelevant information, brings notable performance gains.
#aigraph
https://t.co/8PbQ9LWAyg

[Shuyi @Shuyi817]

SCIBERT is a pretrained language model for scientific text based on BERT which shows significant improvements over BERT in many NLP tasks.
#aigraph
https://t.co/VD9FGwn8Sa

[Shuyi @Shuyi817]

Aspect-Based Sentiment Analysis(ABSA) has been developed to counter the shortcomings of traditional sentiment analysis methods that evaluate a text holistically, attaching a single sentiment label to it.
#aigraph
https://t.co/xqigcOW4oI

[Shuyi @Shuyi817]

We are facing the daunting odds of AI and Quantum Computers in the world of cryptography! Quantum AI threatens to crack codes faster than ever.
#aigraph
https://t.co/ax0YXBcIE9

[Shuyi @Shuyi817]

With InstructDiffusion, the latest computer vision model, we can do tremendous amazing vision tasks with just human instruction!
#aigraph
https://t.co/VDA3HSQaIN

[Shuyi @Shuyi817]

Consensus clustering combines the results of multiple clustering algorithms to produce a more robust and reliable clustering solution that considers the strengths and limitations of each algorithm.
#aigraph
https://t.co/Hcay60FyHW

[Shuyi @Shuyi817]

While generative AI may seem like a shortcut, remember, quality content and genuine connections require human touch. Let's not forget the value of human creativity in the world of content creating.
#aigraph
https://t.co/pIjVbVeJqv

[Shuyi @Shuyi817]

#aigraph Although now no AI satisfies all indicators of consciousness, there is at least one current architecture which meet any one of the most indicators. Will we create a conscious AI one day?
https://t.co/9Tvl9nNMac

[Shuyi @Shuyi817]

#aigraph ClusterLLM, a novel text clustering framework that leverages the power of LLM to improve clustering quality. With ClusterLLM, you can understand the user's preference on clustering through textual instruction or annotated data.
https://t.co/zOd53BVl72

[Shuyi @Shuyi817]

#aigraph FastText, just as its name 'fast' suggests, the biggest advantage is that it can often achieve accuracy comparable to deep networks, but its training time is many orders of magnitude faster than deep networks.
https://t.co/9QTGnEFR3P

[Shuyi @Shuyi817]

#aigraph Compared to traditional text classification, deep learning models enhance performance by improving the presentation learning method, model structure, and additional data and knowledge.
https://t.co/DNuVWHcp49

[Shuyi @Shuyi817]

#aigraph Besides simple and well-known text similarity approaches such as N-gram and cosine similarity, this article also introduce some advanced measures like Normalized Google Distance (NGD), which are worthwhile to explore.
https://t.co/tyH1RTrs55

[Shuyi @Shuyi817]

#aigraph A tutorial about the whole procedure of topic modelling, including data understanding, data preparation, modeling, and evaluation, which can definitely benefit information systems researchers.
https://t.co/a7F3I21aut

[Shuyi @Shuyi817]

#aigraph A very good paper for practitioners and researchers to have a preliminary understanding of different type of clustering algorithms.
https://t.co/DExfucaCnz

[Shuyi @Shuyi817]

#aigraph The paper introduces a family of probabilistic time series models that can analyze the time 'evolution' of topics, which means each year’s articles arise from a set of topics that have evolved from the last year’s topics.
https://t.co/rtF5YxJxJO

[Shuyi @Shuyi817]

#aigraph The Unnatural Instructions dataset, which is generated using language models. Models trained on this dataset outperform that trained on manually-curated datasets, demonstrating the potential of model-generated data as a cost-effective alternative.
https://t.co/OkFBhHifEN

[Shuyi @Shuyi817]

#aigraph The most important part of hierarchical clustering is to calculating the similarity between two clusters to merge or divide the clusters. Different approaches will cause different cluster's shape, and the ability of anti-noise.
https://t.co/aD0RXQkQEw

[Shuyi @Shuyi817]

#aigraph Thanks to Hugging Face, all researchers and end-users are able to access tens of thousands of Transformer-based pre-trained model via its open-source library transformers.
https://t.co/0wyMRe1dKn

[Shuyi @Shuyi817]

#aigraph It shows that there will be a better performence if use dimensionality reduction like UMP before the clustering algorithm, especially those (like HDBSCAN) which don't work well with high-dimension embeddings.
https://t.co/d0LpNXRFri