hammadkhann
MLE @ Flo Health | Previously Cookpad, Alibaba Group, Huawei | MSc Data Science University of Glasgow | IR | NLP | Recommendation systems.
Flo HealthBristol, UK
Pinned Repositories
AI-RecommenderSystem
该仓库尝试整理推荐系统领域的一些经典算法模型
Boolean-Retrieval-Model
The objective of this assignment is to make you understand how different indexes work in retrieving different query from a collection. You will create Inverted index and positional index for a set of collection to facilitate Boolean Model of IR. Inverted files and Positional files are the primary data structure to support the efficient determination of which documents contain specified terms and at which proximity. You also learn to process simple Boolean expression queries through this assignment. Datasets You are given two textual dataset for building inverted and positional index on it. Your dictionary must be written to disk, for each word in the lexicon you must store a file offset to the corresponding posting list, and finally, you should process the raw text collection only once (many real-word collections are so big that the cost of multiple scans is prohibitive). For the Quran dataset, each verse constitutes a document. For the Hadeeth dataset, each hadeeth constitutes a document. Query Processing In this assignment, all you need to implement an information retrieval model called Boolean Information Retrieval Model with some simplified assumptions. You will be provided with a bunch of overly simplified documents collection (corpus), you need to implement a simplified Boolean users queries that can only be formed by joining three terms (t1, t2 and t3) with ( AND, OR and NOT) Boolean operators. For example a user query may be of the form (t1 AND t2 AND t3). For positional queries, the query text contains “/” along with a k intended to return all documents that contains t1 and t2, k words apart. For /0 (zero) means phrase or adjacent terms. Basic Assumption for Boolean Retrieval Model 1. An index term (word) is either present (1) or absent (0) in the document. A dictionary contains all index terms. 2. All index terms provide equal evidence with respect to information needs. ( No frequency count necessary, but in next assignment it can be) 3. Queries are Boolean combinations of index terms at max 3. 4. Boolean Operators (AND, OR and NOT) are allowed. For examples: X AND Y: represents doc that contains both X and Y X OR Y: represents doc that contains either X or Y NOT X: represents the doc that do not contain X 5. Queries of the type X AND Y / 3 represents doc that contains both X and Y and 3 words apart. As we discussed during the lectures, we will implement a Boolean Model by creating a posting list of all the terms present in the documents. You are free to implement a posting list with your choice of data structures; you are only allowed to preprocess the text from the documents in term of tokenization in which you can do case folding and stop-words removal but no-stemming. The stop word list is also provided to you in assignments files. Your query processing routine must address a query parsing, evaluation of the cost, and through executing it to fetch the required list of documents. A command line interface is simply required to demonstrate the working model. You are also provided by a set of 10 queries, for evaluating your implementation. Coding can be done
ColBERT
ColBERT: Contextualized Late Interaction over BERT (SIGIR'20)
E-commerce-Website-for-Electronic-items.
This project deals with the online electronic store to buy some good products easily from home with your secure online payment option. Programming Language: PHP, MySQL, HTML, CSS, JavaScript, PayPal integration.
Effective-CLS-Token-Representation-for-ColBERT-A-Step-Towards-Reducing-the-Storage-Cost
HR-management-System.
Designed a Human Resources Management System (HRMS) software application that combines many human resources functions, including benefits administration, payroll, recruiting and training, and performance analysis and review into one package. I managed and worked in this project as Team Lead. Programming Language: c#,sql server,.net
ML-for-AI-and-DS
Machine Learning for AI and DS Case Studies.
pyterrier_colbert
Sarcasm-Detection-on-Reddit-data.
Python based sarcasm detection using reddit comments dataset.
State-based-LWW-Element-Graph
This repository implements state-based LWW-Element-Graph CRDT in python.
hammadkhann's Repositories
hammadkhann/HR-management-System.
Designed a Human Resources Management System (HRMS) software application that combines many human resources functions, including benefits administration, payroll, recruiting and training, and performance analysis and review into one package. I managed and worked in this project as Team Lead. Programming Language: c#,sql server,.net
hammadkhann/Boolean-Retrieval-Model
The objective of this assignment is to make you understand how different indexes work in retrieving different query from a collection. You will create Inverted index and positional index for a set of collection to facilitate Boolean Model of IR. Inverted files and Positional files are the primary data structure to support the efficient determination of which documents contain specified terms and at which proximity. You also learn to process simple Boolean expression queries through this assignment. Datasets You are given two textual dataset for building inverted and positional index on it. Your dictionary must be written to disk, for each word in the lexicon you must store a file offset to the corresponding posting list, and finally, you should process the raw text collection only once (many real-word collections are so big that the cost of multiple scans is prohibitive). For the Quran dataset, each verse constitutes a document. For the Hadeeth dataset, each hadeeth constitutes a document. Query Processing In this assignment, all you need to implement an information retrieval model called Boolean Information Retrieval Model with some simplified assumptions. You will be provided with a bunch of overly simplified documents collection (corpus), you need to implement a simplified Boolean users queries that can only be formed by joining three terms (t1, t2 and t3) with ( AND, OR and NOT) Boolean operators. For example a user query may be of the form (t1 AND t2 AND t3). For positional queries, the query text contains “/” along with a k intended to return all documents that contains t1 and t2, k words apart. For /0 (zero) means phrase or adjacent terms. Basic Assumption for Boolean Retrieval Model 1. An index term (word) is either present (1) or absent (0) in the document. A dictionary contains all index terms. 2. All index terms provide equal evidence with respect to information needs. ( No frequency count necessary, but in next assignment it can be) 3. Queries are Boolean combinations of index terms at max 3. 4. Boolean Operators (AND, OR and NOT) are allowed. For examples: X AND Y: represents doc that contains both X and Y X OR Y: represents doc that contains either X or Y NOT X: represents the doc that do not contain X 5. Queries of the type X AND Y / 3 represents doc that contains both X and Y and 3 words apart. As we discussed during the lectures, we will implement a Boolean Model by creating a posting list of all the terms present in the documents. You are free to implement a posting list with your choice of data structures; you are only allowed to preprocess the text from the documents in term of tokenization in which you can do case folding and stop-words removal but no-stemming. The stop word list is also provided to you in assignments files. Your query processing routine must address a query parsing, evaluation of the cost, and through executing it to fetch the required list of documents. A command line interface is simply required to demonstrate the working model. You are also provided by a set of 10 queries, for evaluating your implementation. Coding can be done
hammadkhann/AI-RecommenderSystem
该仓库尝试整理推荐系统领域的一些经典算法模型
hammadkhann/Sarcasm-Detection-on-Reddit-data.
Python based sarcasm detection using reddit comments dataset.
hammadkhann/State-based-LWW-Element-Graph
This repository implements state-based LWW-Element-Graph CRDT in python.
hammadkhann/ColBERT
ColBERT: Contextualized Late Interaction over BERT (SIGIR'20)
hammadkhann/Effective-CLS-Token-Representation-for-ColBERT-A-Step-Towards-Reducing-the-Storage-Cost
hammadkhann/ML-for-AI-and-DS
Machine Learning for AI and DS Case Studies.
hammadkhann/pyterrier_colbert
hammadkhann/AI-Maze-Searches
The Maze-Puzzle a rectangular maze where each cell is either empty (contains 0) or blocked (contains 1). There is a special cell designated as start cell and there are some cells designated as goal cells. Our search agent (Raju) need to find an escaping path from given a maze. There are some searching strategies that we need to follow in finding the optimal path. The aims of this maze puzzle is to understand different search strategies. Consider an overly simple example for this programming assignment: Dimension 5 x 5 Initial State= starting cell (0, 0) Goal State= goal cell (4, 4) Optimal Path = {(0, 0), (1, 1), (2, 2), (3, 3), (4, 4)} with Path Cost = 5 when all cell movements have equal unit cost.
hammadkhann/Awesome-Deep-Learning-Papers-for-Search-Recommendation-Advertising
Awesome Deep Learning papers for industrial Search, Recommendation and Advertising. They focus on Embedding, Matching, Ranking (CTR prediction, CVR prediction), Post Ranking, Transfer and Reinforcement Learning.
hammadkhann/awesome-knowledge-graphs
Graph databases, Knowledge Graphs, SPARQ
hammadkhann/Carpooling-App-and-website
Working on the Database for carpool which provide online carpooling enabling students of fast to easily travel with their university mates if they miss their points and good way to earn some money too. Programming Language: SQL, PHP, JAVA
hammadkhann/clothes-in-space
Personalization with deep learning in 100 lines of code
hammadkhann/COIL
NAACL2021 - COIL Contextualized Lexical Retriever
hammadkhann/Condenser
hammadkhann/DeepCTR
Easy-to-use,Modular and Extendible package of deep-learning based CTR models for search and recommendation.
hammadkhann/DeepCTR-Torch
【PyTorch】Easy-to-use,Modular and Extendible package of deep-learning based CTR models.
hammadkhann/EffectivePyTorch
PyTorch tutorials and best practices.
hammadkhann/gpl
Powerful unsupervised domain adaptation method for dense retrieval. Requires only unlabeled corpus and yields massive improvement: https://arxiv.org/abs/2112.07577
hammadkhann/guide
The Student's Guide to @lintool
hammadkhann/hammadkhann
hammadkhann/ISLR-python
An Introduction to Statistical Learning (James, Witten, Hastie, Tibshirani, 2013): Python code
hammadkhann/Jalpc
🍎Jalpc -- A flexible Jekyll theme, 3 steps to build your website.
hammadkhann/Kmeans-Clustering-Doc20.
The task is to cluster a given collection of documents into well-define, identifiable clusters. Clustering is an unsupervised and very challenging problem, here you need to identify many parameters for the task. You are now aware of two basic types of clustering algorithm partition and hierarchical, here you have a choice to apply any of them. The feature choices are still open for you. In order to evaluate the clustering results, you should apply one internal and one external clustering evaluation measure. Dataset The dataset is a subset of famous NEWS20 dataset. It contains 50 textual documents. In supervise learning the input is the only thing available for learning. You can set a baseline for this dataset by using tf*idf based features from the text.
hammadkhann/LLM-Experiments
Experiments related to LLM tech
hammadkhann/machine-learning-exams
This repository contains links to machine learning exams, homework assignments, and exercises that can help you test your understanding.
hammadkhann/MatchPapers
Worth-reading papers and related awesome resources on matching task. 值得一读的匹配任务相关论文与资源集合
hammadkhann/Objectify.
Developed a IOS application for kids using CoreML framework and Mobile Net trained model to detect objects by simply tapping on the screen to take picture of the object our system then tells the name of that object via text and speech representation with surety percentage. I managed and worked in this project as Team Lead. Programming Language: Swift4, CoreML.
hammadkhann/SparrowRecSys
A Deep Learning Recommender System