Here are the slides of a talk I gave at the Data Science Northeast Netherlands Meetup and PyData, where I detail the custom in-house entity linking framework, sentiment analysis, and entity salience scoring model we developed for Company.info (part of FD Mediagroep), in addition to showing some example applications of our corpus of news articles linked to organization profiles.
I’m sharing it here because I think it’s cool, since it’s one of the first project I’ve done at Company.info! Gives you some idea of what we’re working on..
Thursday 19 October, I had the pleasure of hosting the 8th Recommender Systems Amsterdam meetup at FDMG/Company.info. The meetup’s theme was media-content recsys, and we had three talks from industry, dealing with recommending tv programs, music videos, and text articles);
Ghida Ibrahim (Senior Data Scientist, (formerly at) Liberty Global): “Recommender systems for video and TV products”
Bouke Huurnink and Roman Ivanov (XITE): “Music Video Recommendation@XITE”
Robbert van der Pluijm (Head of Bibblio Labs, Bibblio): “Scaling a recommendation service – a threefold story”
And it was a beautiful day. Thanks to everyone who attended my defense, to Daan for this (+ more) great picture, my paranymphs Rutger and Marijn for nymphing like a boss, and my committee for grilling but not burning me. Band pic:
My PhD thesis, Entities of Interest — Discovery in Digital Traces is now available for download. Click on the cover below to head to graus.nu/entities-of-interest and grab your electronic copy of the little booklet that took me 4+ years to write!
On Tuesday May 9th I will participate in a panel discussion on Data & Democracy, which will revolve around the impact of (big) data (mining), profiling, and political micro-targeting on politics and campaigning of the future. Data & Democracy is organized by the Personalised Communication group (a joint effort between UvA’s Communication Science & Information Law groups). See this article (in Dutch) and the flyer (below) for more information!
I was invited to give the opening keynote at the Intelligence Day of the Koninklijke Marechaussee (Military Police) on Big Data and Machine Learning, with the aim to explain the audience what ML and Big Data is.
I spent a disproportionate amount of time on Algorithmic Bias, because I think this is a hugely important topic — in particular for this audience! See the slides of my talk (in Dutch) below, or on slideshare:
Below is an article I wrote with Maarten de Rijke, which was published in nrc.next and NRC Handelsblad under a somewhat misleading title (which wasn’t ours). I cleaned up a Google Translate translation of this article. The translation is far from perfect, but I believe gets the main point across. You can read the original article in Blendle (for €0.29) or on NRC.nl (for free).
A google image search for “three black teens” resulted in mugshot photos, while a search for “three white teens” yielded stock photos of happy smiling youth. Commotion everywhere, and not for the first time. The alleged lack of neutrality of algorithms is a controversial topic. In this controversy, the voice of computer scientists is hardly ever heard. And to have a meaningful discussion on the topic, it is important to understand the underlying technologies.
Our contention, as computer scientists: the lack of neutrality is both necessary and desirable. It is what enables search and recommendation systems to provide us access to huge amounts of information, and let us discover new music or movies. With objective, neutral algorithms, we wouldn’t be able to find anything anymore.
There’s two reasons for this. First, the “usefulness” of information is personal and context-dependent. The quality of a movie recommendation from Netflix, the interestingness of a Facebook post, even the usefulness of a Google search result, varies per person and context. Without contextual information, such as user location, time, or the task performed by the user, even experts do not reach agreement on the usefulness of a search result.
Second, search and recommendation systems have to give us access to enormous quantities of information. Deciding what (not) to display, the filtering of information, is a necessity. The alternative would be a “Facebook” which shows thousands of new messages every single day, making each new visit show a completely new deluge of posts. Or a Netflix which recommends only random movies, so that you can no longer find the movies you really care about.
In short, search and recommendation systems have to be subjective, context-dependent, and adapted to ourselves. They learn this subjectivity and lack of neutrality, from us, their users. The results of these systems are thus a reflection of ourselves, our preferences, attitudes, opinions and behavior. Never an absolute truth.
The idea of an algorithm as a static set of instructions carried out by a machine is misleading. In the context of, for example, Facebook’s news feed, Google’s search results or Netflix’ recommendation, a machine is not told what to do, but told to learn what to do. The systems learn from subjective sources: ourselves, our preferences, our interaction behavior. Learning from subjective sources naturally yields subjective outcomes.
To choose what results to show, a search and recommendation system learns to predict the user’s preferences or taste. To do this, it does what computers do best: counting things. By keeping track of the likes a post receives, or the post’s reading time, the system is able to measure various characteristics of a post. Likes or reading-time are just two examples: in reality, hundreds of attributes are included.
To then learn what is useful for an individual user, the system must determine which features of posts the user considers important. Essential here is to determine the effectiveness of the information displayed. For this, the system gets a goal, such as making sure the user spends more time on the site.
By showing messages with different characteristics (more or less likes, longer or shorter reading times), and to keep track of how long or often the user visits the site, the system can learn which message characteristics makes people spend more time on the website. Things that are simple to measure (clicks, likes, or reading time) are used to bring about more profound changes in user behavior (long term engagement). Furthermore, research has shown that following the personalized recommendations eventually leads to a wider range of choices, and a higher appreciation of the consumed content for users.
The success of modern search and recommendation systems largely results from their lack of neutrality. We should consider these systems as “personalized information intermediaries.” Just like traditional intermediaries (journalists, doctors, opinion leaders), they provide a point of view by filtering and ranking information. And just like traditional intermediaries, it would be wise to seek a second or third opinion when it really matters.
Automated personal assistants such as Google Now, Microsoft Cortana, Siri, M and Echo aid users in productivity-related tasks, e.g., planning, scheduling and reminding tasks or activities. In this paper we study one such feature of Microsoft Cortana: user-created reminders. Reminders are particularly interesting as they represent the tasks that people are likely to forget. Analyzing and better understanding the nature of these tasks could prove useful in inferring the user’s availability, aid in developing systems to automatically terminate ongoing tasks, allocate time for task completion, or pro-actively suggest (follow-up) tasks.
Studying things that people tend to forget has a rich history in the field of social psychology. This type of memory is called “Prospective memory” (or more poetically written: “Remembrance of Things Future“). One challenge in studying PM is that its hard to simulate in a lab study (the hammer of choice for social psychologists). For this reason, most studies of PM have been restricted to “event-based” PM, i.e., memories triggered by an event, modeled in a lab through having someone perform a mundane task, and doing a special thing upon being triggered by an event. Furthermore, the focus in these studies has largely been on retention and retrieval performance of “artificial” memories: subjects were typically given an artificial task to perform. Little is known about the type and nature of actual, real-world, “self-generated” tasks.
Enter Cortana. The user logs we study in this paper represent a rich collection of real-life, actual, self-generated, time-based PM instances, collected in the wild. Studying them in aggregate allows us to better understand the type of tasks that people remind themselves about.
(Yes, sorry, that heading really says big data…) As the loyal reader may have guessed, this paper is the result of my internship at Microsoft Research last summer, and one of the (many) advantages of working at Microsoft Research is the restricted access to big and beautiful data. In this paper we analyze 576,080 reminders, issued by 92,264 people over a period of two months (and we later do prediction experiments on 1.5M+ reminders over a six month time period). Note that this is a filtered set of reminders (a.o. for a smaller geographic area, and we removed all users that only issued a few reminders). Furthermore, when analyzing particular patterns, we filter data to patterns commonly observed across multiple users to study behavior in aggregate and further preserve user privacy: we are not looking at the users behavior at the individual level, but across a large population, to uncover broad and more general patterns. So what do we do to these reminders? The paper consists of three main parts;
1. Task type taxonomy: First, we aim to identify common types of tasks that underlie reminder setting, by studying the most common reminders found in the logs. This analysis is partly data-driven, and partly qualitative; as we are interested in ‘global usage patterns,’ we extract common reminders, defined as reminders that are seen across many users, that contain a common ‘action’ or verb. We do so by identifying the top most common verb phrases (and find 52 verbs that cover ~61% of the reminders in our logs), and proceed by manually labeling them into categories.
2. Temporal patterns: Next, we study temporal patterns of reminders, by looking at correlations between reminder creation and notification, and in temporal patterns for the terms in the reminder descriptions. We study two aspects of these temporal patterns: patterns in when we create and execute reminders (as a proxy to when people typically tend to think about/execute certain tasks), and the duration of the delay between the reminder’s creation and notification (as a proxy to how “far in advance” we tend to plan different things).
3. Predict! Finally, we show how the patterns we identify above generalize, by addressing the task of predicting the day at which a reminder is likely to trigger, given its creation time and the reminder description (i.e., terms). Understanding when people tend to perform certain tasks could be useful for better supporting users in the reminder process, including allocating time for task completion, or pro-actively suggesting reminder notification times, but also for understanding behavior at scale by looking at patterns in reminder types.
As always, no exhaustive summary of the paper point-by-point here, straight into some of our findings (there’s much more in the paper):
We tend to plan for things (i.e., set reminders) at the end of day, and execute them (i.e., reminders trigger) throughout the day, which suggests the end of day is a natural moment for people to reflect upon the tasks that need to be carried out.
The types of things we remind ourselves about are mostly short-term, immediate, tasks such as performing daily chores.
People are more likely to call their mom, and email their dad.
Want to know more? See the taxonomy? See more pretty plots? Look at some equations? Learn how this could improve intelligent assistants? Read the paper!