Building on the experiences of NTCIR-12, there are now four subtasks in the Lifelog task at NTCIR-13.
PHASE 1 TASKS
LAT - Lifelog Annotation (sub) Task
The aim of this subtask is to explore the most effective computer vision algorithms to accurately describe the activities and settings of the individual of lifelog data. An ontology will be generated of important lifelog concepts (activities, environments) and the task will require the development of automated approaches to annotating these concepts. Both image content as well as provided metadata and external evidence sources can be used to generate the annotations. The best performing annotation outputs are encouraged to release the data for other participants to use in the other three following tasks (see below). A subset of the initial ontology of lifelog concepts will be selected for this task. This subset will be represented as topics for evaluation.
The submission for the LAT task is in the form of a single CSV file per run. We can accept a maximum of ten runs per participant. We will release the qrels on the NTCIR website after results are released. The format of the submission file is one line per result in the format as follows:
GROUP_ID, RUN_ID, TOPIC_ID, IMAGE_ID, SCORE (in decreasing order), CONFIDENCE_LEVEL (if no confidence value, include 0)
DCU, DCULAT01, LAT001, u1_2016-08-15_112559_2, 1.0, 0.546352
DCU, DCULAT01, LAT001, u1_2016-08-15_113042_1, 0.98, 0.532413
The submission files should be sent (one email per group) to (ntcir-lifelog@computing.dcu.ie) by the due date with the title ’NTCIR-Lifelog LAT Submission’.
PHASE 2 TASKS
LSAT - Lifelog Semantic Access (sub) Task - same as NTCIR-12
In this sub task, the participants have to retrieve a number of specific moments in a lifelogger's life. We define moments as semantic events, or activities that happened throughout the day. The task can best be compared to a known-item search task.
Example search tasks include:
Find the moment(s) where I was boarding an A380.
Find the moment(s) where I am in my kitchen.
Find the moment(s) where I am playing with my phone.
Find the moment(s) where I am preparing breakfast.
LEST - Event Segmentation (sub) Task
The aim of this subtask is to examine approaches to event segmentation from continual lifelog stream data. Events have always been proposed to be the standard unit of retrieval and there have been a number of suggestions made as to how to segment into events. This task evaluates how well the participants can segment a test set of lifelog data when compared to a human-judgement segmentation generated by the life loggers who gathered the data. To reduce the level of subjectivity in the manual judgement process, the manual segmentation is achieved by segmentation the 90 days of data into fifteen distinct episode types, as follows (label: description):
- traveling: travelling (car, bus, boat, airplane, train, etc)
- f2f: face-to-face interaction with people at home or in the workplace (excluding social interactions)
- computer: using desktop computer / laptop / tablet / smartphone
- cooking: preparing meals (include making tea or coffee) at any location
- eating: eating meals in any location, but not including moments when drinking alone.
- children: taking care of children / playing with children
- home: working in the home (e.g. cleaning, gardening)
- relax: relaxing at home (e.g. TV, having a drink)
- paper: reading any form of paper
- social: socialising outside the home or office
- praying: praying / worshipping / meditating
- shopping: shopping in a physical shop (not online)
- gaming: playing computer games
- physical: physical activities / sports (walking, playing sports, cycling, rowing, etc)
- creative: creative endeavours (writing, art, music)
- other: any other activity not represented by the fourteen labels above.
Submissions are requested that identify each segment begin and end time (to the nearest minute). A sliding window of three minutes will be used to provide a level of fuzzy evaluation.
LIT - Lifelog Insight Task (sub) Task
The aim of this subtask is to gain insights into the lifelogger's life. It follows the idea of the Quantified Self movement that focuses on the visualization of knowledge mined from self-tracking data to provide "self-knowledge through numbers". Participants are requested to generate new types of visualisations and insights about the life of the lifeloggers by generating a themed diary or themed insights related to some target topics. This task is not evaluated in the traditional sense, but participants will be expected to present their work in a special session at NTCIR. We are flexible to the format of the themed diaries. There are five topics defined for this task and they act as guidelines. Participants are free to define their own additional topics for the task. The five core topics are:
- diet: Provide insights into the diet and blood sugar levels of the lifeloggers. For example, how does diet impact on the blood sugar levels of the lifeloggers?
- exercise: Describe the exercise, sleep and physical activities of both lifeloggers
- social: Socialisation levels are a good indicator of the health of individuals. Provide a themed social diary for each individual.
- where: Provide insights onto the location and movement patterns of the lifeloggers (e.g. are they ever in the same place?)
- compare: Comparison between two individuals across multiple dimensions (where the dimension is up to you).