Technical Disclosure Commons Technical Disclosure Commons
Defensive Publications Series
07 Nov 2017
Context based automatic email responses Context based automatic email responses
Victor Carbune
Daniel Keysers
Thomas Deselaers
Follow this and additional works at: https://www.tdcommons.org/dpubs_series
Recommended Citation Recommended Citation
Carbune, Victor; Keysers, Daniel; and Deselaers, Thomas, "Context based automatic email responses",
Technical Disclosure Commons, (November 07, 2017)
https://www.tdcommons.org/dpubs_series/792
This work is licensed under a Creative Commons Attribution 4.0 License.
This Article is brought to you for free and open access by Technical Disclosure Commons. It has been accepted for
inclusion in Defensive Publications Series by an authorized administrator of Technical Disclosure Commons.
Context based automatic email responses
ABSTRACT
When users are not available to respond to emails, they typically set-up automatic
responses to incoming emails using standard messages (e.g., a standard out of office reply) or
simple rule based systems. However, such standard responses may not provide adequate context
and detail to the recipient of the automatic reply. Techniques described here employ machine
learning and heuristic models to dynamically generate contextual responses to emails based on
the communication history between the corresponding email sender and recipient(s).
KEYWORDS
● Scoring mechanism
● Out-of-office
● Automatic response
● E-mail reply
● Inbox
BACKGROUND
Automatic responses are deployed by users when they are out of the office or otherwise
unavailable to respond to email communication. Users set-up automatic responses in advance,
but these responses may become obsolete, as they are independent of the content and context of
future incoming emails. Such static responses may not provide senders of emails adequate
information to take further action.
DESCRIPTION
In situations where users are unable to respond to emails, automatic responses can be
improved by adding contextual information. Such context is identified, for example, from past
2
Carbune et al.: Context based automatic email responses
Published by Technical Disclosure Commons, 2017
communications related to the incoming emails. Techniques described customize automated
responses based on the content of the incoming email and email communications received during
the given automatic response period, e.g., out of office period. The automatically generated
response could be entirely new or may use the static message and then also include appended
content and phrases based on the described techniques.
Fig. 1: Customized automatic email response generator
The automatic response generator (104) uses machine learning techniques to identify,
with user consent and permission, relevant content from incoming emails and match that content
with other communications received by users during the automatic response (e.g., out of office)
period. Based on this matching analysis, machine learning and/or heuristic based techniques are
used to extract meaningful user metadata and content data from previous email communications
(102) and generate relevant automatic email responses (110).
The techniques first determine whether to modify the standard user provided response
based on the incoming email message (“the trigger”) and other emails (“the context”) received
by users during the relevant time period. If a modification is required, customization can be
3
Defensive Publications Series, Art. 792 [2017]
https://www.tdcommons.org/dpubs_series/792
either minimal or extensive, depending on the context. For example, minimal customization may
involve determining relevant details of the person to contact for the inquiry based on the context,
and including this information in the standard automatic response. Alternatively, pertinent
details may be incorporated into the static user provided response where appropriate, for
example, based on topics or keywords in incoming emails. In some cases, extensive
customization involves generating a new free-form text response to the email.
Prediction model(s) (106) may be trained and optimized for different criteria. For
example, machine learning techniques for extracting relevant metadata or content could be
optimized to identify private and sensitive information and exclude such content from the
automatic response to preserve user privacy. Similarly, matching can be optimized based on
semantic analysis and clustering (e.g., of different topics) so that the incoming emails are
appropriately matched with relevant email communication previously received by the user.
To avoid sharing potentially sensitive data, users are enabled to configure privacy
settings in advance. For example, a default privacy level is tailored such that content included in
the automatic reply does not include sensitive data. A user interface (108) is provided to easily
customize and turn on and off privacy options. For example, the present techniques can be either
disabled completely or selectively disabled, e.g., for internal versus external emails and
customized, turned off for particular groups, etc. Also, the user interface allows automatic
replies to be turned off.
For emails corresponding to sensitive matters, only contact information of the relevant
people who can respond to the email sender is included in the automatic reply while in more
permissive contexts, additional details could be incorporated. These additional details in the
context of a project may include relevant events related to the project (conference/meeting),
4
Carbune et al.: Context based automatic email responses
Published by Technical Disclosure Commons, 2017
pertinent project contacts/teams and project status updates (e.g., started, in progress, on hold,
cancelled).
For example, where sensitive topics are involved in the email, the automatic reply may
say: “Consider contacting X, Y, Z for the latest updates on this project.” However, in a more
permissive context, the automatic reply can say: “Appears that X just began working on the
project. Consider grabbing a coffee with X until I return, when I’ll be happy to chat with you.”
Similarly, automatic response to an email (without sensitive content) inquiring about a meeting
may say: “Contact Y for additional information regarding meeting ABC or to join the meeting.”
Also, the relationship of the email sender to the project may inform the automatic reply.
For example, if the sender has been deeply involved with the project (based on past emails), a
detailed response customized for the sender may be suitable. However, if an incoming email is
the initial communication regarding a project, a brief introductory response may be appropriate.
Automatic emails may also be sent to project stakeholders alerting them about an email
sender with interest in the project: “Looks like Bob is interested in the project. If you have time,
consider chatting with him.” This interest could be in part determined by the number of emails
that Bob sent about the project, for example, Bob may have sent fifty emails about the project.
Also, incoming emails that are urgent or need user approval may be escalated to relevant
stakeholders to enable prompt action before the user’s return or subsequent reply.
Examples of Use
Alice sets up an automatic email response for all email communication received during
her vacation from work. While on vacation, Alice may receive the below message from Bob.
5
Defensive Publications Series, Art. 792 [2017]
https://www.tdcommons.org/dpubs_series/792
Fig. 2: Sample Incoming Mail
Standard automatic responses are typically static predetermined messages such as Fig. 3.
Fig. 3: Sample Standard Automatic Response
However, the present techniques offer a more useful response such as the message shown
in Fig. 4.
Fig. 4: Sample Customized Automatic Response
The techniques can incorporate related features to enhance the usability of the automatic
response generator. For example, translation techniques can be employed if a user receives or
sends emails in different languages (e.g., language used for communication within the team is
6
Carbune et al.: Context based automatic email responses
Published by Technical Disclosure Commons, 2017
different from language used for communication with the customer). Also, content analysis
techniques may be used to decipher and extract content from videos, audios, images, etc.
Other machine learning based features can be incorporated. For example, emails
received during the automatic response period may be prioritized for user review using a scoring
mechanism to rate the effectiveness of automatic responses generated.
In situations in which certain implementations discussed herein may collect or use
personal information about users (e.g., user data, information about a user’s social network,
user's location and time at the location, user's biometric information, user's activities and
demographic information), users are provided with one or more opportunities to control whether
information is collected, whether the personal information is stored, whether the personal
information is used, and how the information is collected about the user, stored and used. That
is, the techniques discussed herein collect, store and/or use user personal information specifically
upon receiving explicit authorization from the relevant users to do so.
For example, a user is provided with control over whether programs or features collect
user information about that particular user or other users relevant to the program or feature.
Each user for which personal information is to be collected is presented with one or more options
to allow control over the information collection relevant to that user, to provide permission or
authorization as to whether the information is collected and as to which portions of the
information are to be collected. For example, users can be provided with one or more such
control options over a communication network. In addition, certain data may be treated in one or
more ways before it is stored or used so that personally identifiable information is removed. As
one example, a user’s identity may be treated so that no personally identifiable information can
7
Defensive Publications Series, Art. 792 [2017]
https://www.tdcommons.org/dpubs_series/792
be determined. As another example, a user’s geographic location may be generalized to a larger
region so that the user's particular location cannot be determined.
CONCLUSION
Techniques described generate automatic email responses that are tailored to the content
of the incoming emails. After securing user permission, incoming email content is matched with
context from past email communications to identify and extract details to be included in the
automatically generated response. The automatically generated response can be entirely new or
can use the static message and also include appended content and phrases based on the described
techniques. Sensitive data is protected by allowing users to specify privacy levels with respect to
content, context and contacts. The described techniques have many applications and may be
used by, for example, email providers/clients, automatic email response services as well as
email/digital assistants and operating systems.
8
Carbune et al.: Context based automatic email responses
Published by Technical Disclosure Commons, 2017
