December 2022 Summaries

Hex has introduced new features for charting, including interactive visual filtering, an all-new editor UI, and custom color palettes. Visual filtering allows users to select data points of interest directly on the charts, making it easier to explore and analyze data visually. The redesigned chart cell provides a more intuitive interface with finer-grained styling options. Additionally, admins can configure custom color palettes for branded charts. These improvements aim to make charting more accessible and efficient in Hex's analytics workflow.

Text preprocessing is an essential step in preparing text data for natural language processing (NLP) tasks. It involves a series of techniques aimed at reducing noise in the dataset while retaining relevant information. Key steps include tokenization, normalization, removing unwanted characters and stop words, lemmatization, and stemming. These methods help to simplify text, reduce vocabulary size, and improve model performance on NLP tasks.

Sentiment analysis is a natural language processing technique used to detect or extract the tone or feeling from a body of text, usually classifying it as positive, negative, or neutral. It can also be used to classify emotions, intentions, and predict user ratings. In this tutorial, we learn how to use Huggingface's pre-trained sentiment analysis models for analyzing customer reviews. We first load the dataset containing cell phone reviews from Kaggle and perform some data cleaning and preprocessing. Then, we pass our list of reviews to a BERT model finetuned for sentiment analysis tasks to obtain predicted ratings. Finally, we calculate an accuracy score to measure the model's performance in predicting actual review ratings and overall sentiment. The tutorial also demonstrates how to create a report showcasing the percentage of each sentiment category for all reviews for each company and a word cloud visualization to understand common words associated with different sentiments.

Exploratory Data Analysis (EDA) is an approach to data analysis that focuses on understanding the structure of your data. It involves visualizing and summarizing the data to gain insights into its characteristics and relationships between variables. EDA helps in selecting appropriate analytical techniques, models, and interpreting their outputs accurately. The process includes checking the size of the dataset, column names, data types, missing values, and descriptive statistics. Data visualization libraries such as Seaborn can be used to create plots like boxplots, scatter plots, histograms, and heatmaps for better understanding of the data. EDA is crucial before moving into more sophisticated modeling techniques, as it helps in identifying potential issues or biases within the dataset.

Human Interest's data team has successfully implemented self-service analytics products to empower stakeholders across the company to make smarter, data-driven decisions. By leveraging Hex for their day-to-day work, they have created a process that allows more technical users to asynchronously augment the abilities of less technical users. The success of self-service analytics relies on making data products easy to find and fitting the mental models for search that people have. Simplicity is key in ensuring general users can autonomously use the self-service app, with documentation and consistent naming conventions playing important roles. Hex has been a core part of their infrastructure powering self-service analytics, allowing stakeholders to access and leverage data independently without interfacing with the data organization. The adoption of self-service can empower both stakeholders and analysts, enabling them to spend more time working together and focusing on answering the most important questions.

Nikhil Sabharwal from Human Interest describes the process of developing a self-service analytics infrastructure within the company to empower non-technical stakeholders to make data-driven decisions. By using tools like Hex, the data team has created an environment where employees can independently access and analyze data without needing extensive technical skills or direct interaction with data scientists. This approach involves creating self-service analytics products that simplify data access through user-friendly interfaces and formats, such as spreadsheets and simple data visuals, tailored to the preferences of different stakeholders. The integration of tools like Hex facilitates this process by enabling SQL generation and making data science models accessible, thereby allowing teams to focus on more complex tasks and fostering a more consultative relationship between analysts and stakeholders. The journey toward self-service analytics is continuous, requiring constant adaptation and learning to meet the evolving needs of the organization.